Helicobacter pylori detection method based on mass-spectrometry identification

By employing a strategy based on visual pattern feature extraction and query network matching, and utilizing a dilated convolutional neural network model to construct a fingerprint spectrum visual pattern feature database, the problem of insufficient sensitivity and specificity of traditional mass spectrometry identification methods in Helicobacter pylori detection is solved, thus achieving more reliable mass spectrometry identification of Helicobacter pylori.

WO2026237192A1PCT designated stage Publication Date: 2026-11-19LANZHOU BAIYUAN GENE TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/CN2025/113405
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-05-15
Filing Date
2025-08-08
Publication Date
2026-11-19

AI Technical Summary

Technical Problem

Existing mass spectrometry-based methods for detecting Helicobacter pylori have shortcomings in terms of detection sensitivity and specificity. In particular, they are prone to identification confusion or failure when processing low-quality spectra or distinguishing closely related species. Traditional peak matching or overall spectrum similarity calculation methods are difficult to cope with spectrum drift caused by subtle differences between Helicobacter pylori strains and instrument conditions.

Method used

A strategy based on visual pattern feature extraction and query network matching is adopted. Deep pattern information of mass spectrometry data is extracted through a dilated convolutional neural network model to construct a fingerprint spectrum visual pattern feature database. Bacterial species detection is performed by using fingerprint spectrum pattern feature query response analysis to achieve more reliable Helicobacter pylori mass spectrometry identification.

Benefits of technology

It improves the accuracy and robustness of Helicobacter pylori detection, enabling more stable identification of Helicobacter pylori, resisting interference from noise and subtle changes, and enhancing the reliability and precision of detection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN2025113405_19112026_PF_FP_ABST
    Figure CN2025113405_19112026_PF_FP_ABST
Patent Text Reader

Abstract

A Helicobacter pylori detection method based on mass-spectrometry identification. The method comprises: first, performing MALDI mass-spectrometry acquisition on a sample to be tested, so as to obtain a protein fingerprint spectrum; then, performing preprocessing by means of denoising and normalization processing, so as to further extract a discriminative visual-pattern feature to be tested, and extracting reference visual-pattern features of known bacterial species from a constructed bacterial-species reference database; and finally, introducing feature-domain-based query-response analysis for protein-fingerprint-spectrum pattern features to learn complex matching relationships between the feature to be tested and the reference features, so as to obtain a query-response coded representation of the fingerprint-spectrum visual-pattern feature, and finally determining a bacterial species detection result on the basis of a query response. The strategy based on visual-pattern feature extraction and query-network-based matching aims to make full use of deep-level pattern information of mass-spectrometry data, thereby achieving more reliable mass-spectrometry identification of Helicobacter pylori.
Need to check novelty before this filing date? Find Prior Art

Description

A Mass Spectrometry-Based Method for Detecting Helicobacter pylori Technical Field

[0001] This application relates to the field of Helicobacter pylori detection technology, and more specifically, to a Helicobacter pylori detection method based on mass spectrometry identification. Background Technology

[0002] Helicobacter pylori is a common Gram-negative bacillus that primarily colonizes the human gastric mucosa. It is a major causative agent of various upper gastrointestinal diseases, including chronic gastritis, peptic ulcers, gastric mucosa-associated lymphoid tissue lymphoma, and even gastric cancer. Given its widespread infection rate and serious health consequences, developing rapid, accurate, and reliable detection methods for Helicobacter pylori is of crucial clinical and public health value for early diagnosis, effective treatment, and prevention of serious complications such as gastric cancer.

[0003] In recent years, matrix-assisted laser desorption / ionization time-of-flight mass spectrometry (MALDI-TOF MS) has been widely used in the field of microbial identification due to its speed, high throughput, and relatively low cost. Existing mass spectrometry-based microbial detection schemes typically acquire the protein fingerprint spectrum of the target microorganism (either cultured or directly enriched from the sample) and then compare the resulting list of mass spectra or the entire spectrum with reference spectra of known bacterial species stored in a database to determine the species of the target microorganism. However, when applied to the detection of specific bacterial species such as Helicobacter pylori, these traditional methods based on peak matching or overall spectrum similarity calculation have certain limitations. For example, subtle differences in protein expression may exist between different Helicobacter pylori strains or with other symbiotic or pathogenic bacteria, resulting in mass spectra that are similar overall but have different key characteristic peaks. At the same time, factors such as sample preparation, culture conditions, and instrument status may also cause spectral drift and peak intensity fluctuations. These changes may interfere with the accuracy of traditional comparison algorithms, especially when processing low-quality spectra or distinguishing closely related species, which can easily lead to identification confusion or failure, affecting the sensitivity and specificity of detection.

[0004] Therefore, an optimized mass spectrometry-based detection scheme for Helicobacter pylori is desired. Summary of the Invention

[0005] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a Helicobacter pylori detection method based on mass spectrometry identification. This method employs a strategy based on visual pattern feature extraction and query network matching, aiming to fully utilize the deep pattern information of mass spectrometry data to achieve more reliable Helicobacter pylori mass spectrometry identification.

[0006] According to one aspect of this application, a method for detecting Helicobacter pylori based on mass spectrometry is provided, comprising: mixing a sample to be tested with a MALDI matrix and spotting the mixture onto a MALDI target plate, and acquiring a protein fingerprint spectrum using a MALDI-TOF mass spectrometer; extracting a set of protein reference fingerprint spectra for known bacterial species from a bacterial species reference database; preprocessing the protein fingerprint spectrum to obtain a preprocessed protein fingerprint spectrum; extracting fingerprint spectrum visual pattern features from the preprocessed protein fingerprint spectrum to obtain visual pattern features of the protein fingerprint spectrum to be tested; extracting fingerprint spectrum visual pattern features from the set of protein reference fingerprint spectra for known bacterial species to obtain a set of protein fingerprint spectrum reference visual pattern features; performing a fingerprint spectrum pattern feature query on the visual pattern features of the protein fingerprint spectrum to be tested and the set of protein fingerprint spectrum reference visual pattern features to obtain fingerprint spectrum visual pattern feature query response features; and determining the bacterial species detection result based on the fingerprint spectrum visual pattern feature query response features.

[0007] In one possible implementation, the preprocessing includes baseline correction, noise smoothing, peak identification, and alignment.

[0008] In one possible implementation, extracting fingerprint visual pattern features from the preprocessed protein fingerprint spectrum to obtain the visual pattern features of the protein fingerprint spectrum to be tested includes: passing the preprocessed protein fingerprint spectrum through a fingerprint visual pattern feature extractor based on a dilated convolutional neural network model to obtain a visual pattern feature encoding vector of the protein fingerprint spectrum to be tested as the visual pattern features of the protein fingerprint spectrum to be tested.

[0009] In one possible implementation, extracting fingerprint visual pattern features from the set of protein reference fingerprint spectra of the known bacterial species to obtain a set of protein fingerprint reference visual pattern features includes: passing the protein reference fingerprint spectra of each known bacterial species in the set of protein reference fingerprint spectra of the known bacterial species through the fingerprint visual pattern feature extractor based on the dilated convolutional neural network model to obtain a set of protein fingerprint reference visual pattern feature encoding vectors, which is then used as the set of protein fingerprint reference visual pattern features.

[0010] In one possible implementation, performing fingerprint pattern feature query on the set of visual pattern features of the protein fingerprint spectrum to be tested and the set of reference visual pattern features of the protein fingerprint spectrum to obtain fingerprint pattern visual pattern feature query response features includes: enhancing the visual pattern feature encoding vector of the protein fingerprint spectrum to be tested, and then performing individual semantic query encoding on each of the reference visual pattern feature encoding vectors of the protein fingerprint spectrum in the set of reference visual pattern feature encoding vectors to obtain a set of individual semantic query score encoding vectors of the visual pattern of the protein fingerprint spectrum to be tested; calculating the individual semantic weight of each individual semantic query score encoding vector of the visual pattern of the protein fingerprint spectrum to be tested in the set of individual semantic query score encoding vectors to be tested to obtain a set of individual query semantic self-attention weights of the visual pattern of the protein fingerprint spectrum to be tested; and aggregating the set of individual semantic query score encoding vectors of the visual pattern of the protein fingerprint spectrum to be tested based on the set of individual query semantic self-attention weights of the visual pattern of the visual pattern of the protein fingerprint spectrum to obtain the fingerprint pattern visual pattern feature query response encoding vector.

[0011] In one possible implementation, after feature enhancement of the visual pattern feature encoding vector of the protein fingerprint spectrum to be tested, it is then subjected to individual semantic query encoding with each of the protein fingerprint spectrum reference visual pattern feature encoding vectors in the set of protein fingerprint spectrum reference visual pattern feature encoding vectors to obtain a set of individual semantic query score encoding vectors for the visual pattern of the protein fingerprint spectrum to be tested. This includes: performing feature enhancement based on deconvolution encoding on the visual pattern feature encoding vector of the protein fingerprint spectrum to be tested to obtain an enhanced visual pattern feature encoding vector of the protein fingerprint spectrum to be tested, wherein the enhanced visual pattern feature encoding vector of the protein fingerprint spectrum to be tested has the same feature scale as each of the protein fingerprint spectrum reference visual pattern feature encoding vectors in the set of protein fingerprint spectrum reference visual pattern feature encoding vectors; and performing individual semantic query encoding on the enhanced visual pattern feature encoding vector of the protein fingerprint spectrum to be tested and each of the protein fingerprint spectrum reference visual pattern feature encoding vectors in the set of protein fingerprint spectrum reference visual pattern feature encoding vectors to obtain a set of individual semantic query score encoding vectors for the visual pattern of the protein fingerprint spectrum to be tested.

[0012] In one possible implementation, the individual semantic weights of each individual semantic query score encoding vector in the set of individual semantic query score encoding vectors of the visual pattern of the protein fingerprint spectrum to be tested are calculated to obtain a set of self-attention weights for individual semantic queries of the visual pattern of the protein fingerprint spectrum to be tested. This includes: determining the individual semantic matching degree of each individual semantic query score encoding vector in the set of individual semantic query score encoding vectors of the visual pattern of the protein fingerprint spectrum to be tested based on the self-distribution characteristics of the feature set of the set of individual semantic query score encoding vectors of the visual pattern of the protein fingerprint spectrum to be tested, to obtain a set of individual semantic matching degrees for the visual pattern of the protein fingerprint spectrum to be tested; and inputting the set of individual semantic matching degrees for the visual pattern of the protein fingerprint spectrum to be tested into a relational gating proxy module to obtain a set of self-attention weights for individual semantic queries of the visual pattern of the protein fingerprint spectrum to be tested.

[0013] In one possible implementation, based on the self-distribution characteristics of the feature set of the set of visual pattern semantic query score encoding vectors of the protein fingerprint spectrum to be tested, the unit semantic matching degree of each unit semantic query score encoding vector of the visual pattern semantic query score encoding vector of the protein fingerprint spectrum to be tested is determined to obtain the set of unit semantic matching degrees of the visual pattern semantic query score encoding vector of the protein fingerprint spectrum to be tested. This includes: performing inherent alignment calibration on each unit semantic query score encoding vector of the visual pattern semantic query score encoding vector of the protein fingerprint spectrum to be tested in the set of visual pattern semantic query score encoding vectors to be tested to obtain the set of calibrated unit semantic query score encoding vectors of the visual pattern semantic query score encoding vector of the protein fingerprint spectrum to be tested; and calculating the unit semantic matching degree of each calibrated unit semantic query score encoding vector of the visual pattern ...

[0014] In one possible implementation, determining the bacterial species detection result based on the fingerprint spectrum visual pattern feature query response feature includes: passing the fingerprint spectrum visual pattern feature query response encoding vector through a classifier-based bacterial species detector to obtain the bacterial species detection result, wherein the bacterial species detection result is used to represent the bacterial species category label.

[0015] This application provides a method for detecting Helicobacter pylori based on mass spectrometry. First, the sample is acquired using MALDI mass spectrometry to obtain its protein fingerprint. Then, noise removal and standardization are performed through preprocessing to further extract discriminative visual pattern features. Simultaneously, reference visual pattern features for each known bacterial species are extracted from a pre-constructed bacterial species reference database. Finally, a feature domain-based protein fingerprint pattern feature query response analysis is introduced to learn the complex matching relationship between the target feature and the reference features, resulting in an encoded representation of the fingerprint visual pattern feature query response. Based on this query response, the bacterial species detection result is finally determined. This strategy, based on visual pattern feature extraction and query network matching, aims to fully utilize the deep pattern information of mass spectrometry data to achieve more reliable Helicobacter pylori mass spectrometry identification. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings of the embodiments of this application will be briefly described below. Obviously, the drawings described below only relate to some embodiments of this application, and are not intended to limit this application.

[0017] Figure 1 is a schematic flowchart of the Helicobacter pylori detection method based on mass spectrometry identification according to an embodiment of this application.

[0018] Figure 2 is a schematic diagram of data flow in the Helicobacter pylori detection method based on mass spectrometry identification according to an embodiment of this application.

[0019] Figure 3 is a schematic flowchart of step S6 in the Helicobacter pylori detection method based on mass spectrometry identification according to an embodiment of this application.

[0020] Figure 4 is a schematic flowchart of step S61 in the Helicobacter pylori detection method based on mass spectrometry identification according to an embodiment of this application.

[0021] Figure 5 is a schematic flowchart of step S62 in the Helicobacter pylori detection method based on mass spectrometry identification according to an embodiment of this application. Detailed Implementation

[0022] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are also within the scope of protection of this application.

[0023] To address the aforementioned technical problems, this application proposes a mass spectrometry-based method for detecting Helicobacter pylori, aiming to improve the accuracy and robustness of mass spectrometry detection of Helicobacter pylori. Its core lies in moving beyond reliance solely on discrete peak positions and intensities, focusing instead on extracting visual pattern features from the protein fingerprint spectrum after preprocessing such as baseline correction, noise smoothing, peak identification, and alignment. These features better reflect the overall morphology and structure of the spectrum. In other words, the protein fingerprint spectrum, like an image, contains rich pattern information. Extracting its visual pattern features captures the macroscopic structure and key region feature distribution of the spectrum. These features may be more resistant to noise and subtle changes than the presence or precise location of a single peak, thus more stably representing the bacterial species. Correspondingly, the same operation is performed on the reference fingerprint spectra of known bacterial species in the bacterial species reference database to extract reference visual pattern features and construct a reference database based on visual pattern features. Furthermore, to achieve efficient and accurate matching, this solution introduces feature domain-based protein fingerprint spectrum pattern feature query response analysis to learn the complex matching relationship between the target feature and the reference feature. This query network goes beyond simple distance calculations or similarity scoring, intelligently determining which reference bacterial species' visual pattern is most similar to that of the test sample. It then outputs a fingerprint-like visual pattern feature query response encoding, and based on this response, ultimately determines the bacterial species detection result. This strategy, based on visual pattern feature extraction and query network matching, aims to fully utilize the deep pattern information in mass spectrometry data to achieve more reliable Helicobacter pylori mass spectrometry identification.

[0024] To address the aforementioned technical problems, this application proposes a method for detecting Helicobacter pylori based on mass spectrometry identification. Figure 1 is a schematic flowchart of the Helicobacter pylori detection method based on mass spectrometry identification according to an embodiment of this application, and Figure 2 is a schematic diagram of data flow in the Helicobacter pylori detection method based on mass spectrometry identification according to an embodiment of this application. As shown in Figures 1 and 2, the Helicobacter pylori detection method based on mass spectrometry identification includes: S1, mixing the sample to be tested with a MALDI matrix and spotting it onto a MALDI target plate, and acquiring a protein fingerprint spectrum using a MALDI-TOF mass spectrometer; S2, extracting a set of protein reference fingerprint spectra of known bacterial species from a bacterial species reference database; S3, preprocessing the protein fingerprint spectrum to obtain a preprocessed... S4, Extracting visual pattern features from the preprocessed protein fingerprint spectrum to obtain the visual pattern features of the protein fingerprint spectrum to be tested; S5, Extracting visual pattern features from the set of protein reference fingerprint spectra of known bacterial species to obtain the set of protein fingerprint spectrum reference visual pattern features; S6, Performing fingerprint pattern feature query on the set of the visual pattern features of the protein fingerprint spectrum to be tested and the set of the protein fingerprint spectrum reference visual pattern features to obtain the fingerprint pattern feature query response features; S7, Determining the bacterial species detection result based on the fingerprint pattern feature query response features.

[0025] Specifically, in step S1, the sample to be tested is mixed with the MALDI matrix and spotted onto a MALDI target plate, and a protein fingerprint spectrum is acquired using a MALDI-TOF mass spectrometer. It should be understood that in proteomics analysis, the MALDI matrix (such as α-cyano-4-hydroxycinnamic acid) can effectively promote the ionization of peptides or proteins in the sample, possessing the essential function of suppressing interference and enhancing the signal. Suspending the biological sample to be tested (such as wet cells of cultured Helicobacter pylori) at a specified concentration in a suitable solution, and then thoroughly mixing it with the MALDI matrix prepared in a specific ratio, ensures sufficient contact between the protein and the matrix crystals. Subsequently, an appropriate amount of the mixture (e.g., 1 μL) is dispensed into the designated wells of the MALDI target plate and allowed to evaporate naturally at room temperature, so that the protein / peptide is uniformly embedded in the matrix crystals. This solid-phase spotting significantly improves the signal-to-noise ratio and spectral reproducibility of subsequent analyses. After sample preparation, the MALDI target plate is loaded into the MALDI-TOF mass spectrometer. The MALDI-TOF mass spectrometer scans samples point-by-point according to preset parameters (such as laser energy and pulse frequency). Laser pulses irradiate the mixing points of the sample matrix, and the matrix molecules absorb the excitation energy, desorbing and ionizing the proteins / peptides embedded within, thus obtaining the time-of-flight distribution information of charged ions. The spectra are calibrated internally and recorded by an external data acquisition system, ultimately outputting a fingerprint mass spectrum for each sample, typically presented as a two-dimensional spectral line of m / z (mass-to-charge ratio) and intensity. For example, Helicobacter pylori, after specific culture, is collected by centrifugation, diluted, and then mixed with MALDI matrix solution at a 1:1 ratio. One microliter of the mixture is precisely added dropwise to the designated wells of a MALDI target plate, allowed to air dry to form a uniform film, and then loaded onto the target plate. Protein fingerprint spectra are then collected step-by-step at preset target points.

[0026] Specifically, in step S2, a set of protein reference fingerprint spectra for known bacterial species is extracted from the bacterial species reference database. It should be understood that in clinical or actual microbiological testing scenarios, Helicobacter pylori not only exhibits diverse strain profiles but is also easily affected by factors such as sample processing, culture conditions, and experimental instruments, leading to variations in protein fingerprint spectra across different batches or laboratories, including spectral drift, peak intensity fluctuations, and even the disappearance and addition of characteristic peaks. Traditional comparison algorithms based on peak values ​​or simple full-spectrum similarity, relying on discrete peak information and surface feature distribution, struggle to address the potentially complex variations and subtle differences in protein spectra between Helicobacter pylori and its closely related species. Therefore, in this application, the set of protein fingerprint spectra for known bacterial species (including standard strains of Helicobacter pylori at various levels) is standardized and incorporated into the reference database.

[0027] Specifically, in step S3, the protein fingerprint spectrum is preprocessed to obtain a preprocessed protein fingerprint spectrum. It should be understood that raw MALDI-TOF mass spectrometry data is often affected by various factors, such as sample preparation, culture conditions, and instrument status, which may cause baseline drift, contain a large amount of random noise, exhibit fluctuations in peak intensity, and slight shifts in peak position (i.e., spectral drift). These defects in the raw data can severely interfere with subsequent analysis. Both traditional peak-matching methods and methods based on visual pattern features require relatively pure and standardized data input. Specifically, in the technical solution of this application, the preprocessing includes baseline correction, noise smoothing, peak identification, and alignment processing. Baseline correction aims to eliminate background signals caused by slow changes in the matrix or other non-peptide substances, ensuring that peak heights accurately reflect peptide abundance. Noise smoothing aims to reduce high-frequency random fluctuations in the spectrum, improve the signal-to-noise ratio, and make true protein peaks easier to detect. Peak identification (or peak extraction) aims to accurately identify characteristic peaks representing protein or peptide fragments from the smoothed spectrum, determining their mass-to-charge ratio position and intensity information. Alignment processing is used to correct systematic shifts in peak positions caused by minor instrument differences or calibration issues between different samples or multiple measurements of the same sample, ensuring that peaks of the same biomarker appear in comparable positions in different spectra.

[0028] Specifically, baseline correction, as the first step in mass spectrometry data processing, aims to eliminate background interference signals in the spectrum. These background signals primarily originate from matrix residues, ion suppression effects, and the accumulation of non-specific ions during analysis, leading to a "false base" phenomenon—a shift or even slow fluctuations in the low-frequency region. Baseline correction algorithms include rolling ball, TopHat filtering, and least squares fitting. The rolling ball algorithm slides along the laser time axis with a specific window size to simulate the lower bound of the background curve, thus achieving baseline removal. For example, with a window range of 200 Da around the m / z spectral line, the baseline is fitted by taking local minima close to the spectral line using the simulated lower edge of the rolling ball, and then the baseline profile is subtracted point by point from the original signal to restore the true peak height. TopHat filtering suppresses slowly changing components in the spectrum by constructing morphological structural elements. The least squares method uses polynomial fitting to form a background parabola, which is then used to correct the background function by differencing the original spectrum. High-quality baseline correction ensures that the signal intensity reflects only the true protein peak concentration, effectively suppressing interference from spurious signals.

[0029] After baseline correction, further noise smoothing is required for the remaining signal. High-frequency noise in mass spectrometry data mainly manifests as random spikes, noise bands, and periodic micro-fluctuations. Effective noise smoothing should preserve the sharpness and contrast of the true peak shape while eliminating non-biologically relevant signal fluctuations. Commonly used noise smoothing methods include moving average, Gaussian filtering, and Savitzky-Golay smoothing. Moving average calculates the mean by sliding within a fixed window (e.g., 5–11 points), suitable for high-frequency noise reduction but with a slight reduction in resolution. Gaussian filtering uses a Gaussian kernel to weight the original signal, filtering out abrupt noise while ensuring peak shape integrity. Savitzky-Golay filtering uses a low-order polynomial to fit the signal within a local sliding window, balancing noise suppression and peak shape detail preservation. Taking MALDI-TOF proteomics data as an example, the window width is typically set to 7–15 points, and the filter order to 2–3. The parameters need to be adjusted based on a balance between specific spectral quality and expected resolution. Through multiple experiments and empirical evaluations, a standardized signal trajectory with a clear peak structure, low noise background, and natural and smooth peak shape was finally obtained.

[0030] Following the noise-smoothed spectrum, peak identification processing is required to accurately locate protein / peptide characteristic peaks and extract their key information parameters (such as m / z value, peak height, peak area, and peak width). The peak identification process needs to be dynamically and adaptively adjusted based on the overall signal-to-noise ratio, local noise level, and peak spacing of the spectrum. Commonly used methods include threshold determination based on local maxima, peak shape discrimination based on first and second derivatives, and multi-scale detection combined with prior biological information. In practice, the spectrum data points are traversed using a sliding window approach. For each data point, the difference between its neighboring points is calculated. If the value of a point is higher than its left and right nearest neighbors and reaches a preset signal-to-noise ratio threshold (e.g., S / N > 3), and its locally fitted peak width and peak area satisfy the protein peak shape distribution characteristics, then the point can be determined as a legitimate peak. For complex samples (such as Helicobacter pylori proteome spectra), signal processing techniques such as breakpoint detection and continuous peak analysis can be introduced to merge weak and false peaks from multiple detections, improving the accuracy of peak identification. Peak identification parameters such as minimum peak height, minimum peak width, and minimum peak area need to be set flexibly according to the performance of the mass spectrometer and the specific experimental system.

[0031] After accurate peak identification, multiple measurements across batches of samples or on a single sample often result in slight shifts in the x-axis (m / z dimension) of the same protein peak due to ion source energy drift, time base adjustment errors, etc. To ensure effective comparison of peak information between spectra and between the reference database and the sample being tested, peak alignment processing is required for all spectra. Peak alignment methods include global linear mapping, local nonlinear registration, and marker-based dynamic time normalization. Linear mapping assumes that the overall spectral drift is constant and uses univariate linear regression to assist in calibrating the entire spectrum by calculating the m / z deviation of the internal standard peak or strong characteristic peak in different samples. For complex drift patterns, nonlinear registration can be used. First, a set of co-occurring peaks (i.e., homologous marker peak groups) is identified in each spectrum, and then the m / z sequence is matched using a dynamic time warping algorithm to achieve non-equidistant global and local mapping. In addition, existing advanced spectrum alignment toolkits all provide automated batch alignment of multiple spectra. Their underlying technology can use interpolation weighting, time-domain normalization, waveform hierarchical clustering and other techniques to ensure that the correspondence of each identified peak is consistent in all samples or all database reference spectra.

[0032] In one specific embodiment, after acquiring the protein fingerprint spectrum, the original profile-type data is first loaded into professional data processing software, and a rolling sphere baseline correction with a window size of 15 Da is applied. Subsequently, Savitzky-Golay filtering with a width of 7 points is used for noise reduction, and effective peaks are identified by maxima detection with a signal-to-noise ratio threshold of 3. Then, 3-5 internal standard peaks are selected for linear / nonlinear alignment mapping to obtain the preprocessed protein fingerprint spectrum.

[0033] Specifically, in step S4, visual pattern features of the fingerprint spectrum are extracted from the preprocessed protein fingerprint spectrum to obtain the visual pattern features of the fingerprint spectrum of the protein to be tested. It should be understood that traditional Helicobacter pylori detection methods rely excessively on discrete peak positions, peak intensities, or overall spectral similarity, making them susceptible to spectral drift, intensity fluctuations, and subtle differences between strains caused by factors such as sample preparation and instrument status, resulting in insufficient accuracy and stability in identification. Therefore, to move beyond these volatile information and instead uncover fingerprint spectrum visual pattern features that better reflect the stable internal structure of the spectrum, the technical solution of this application further extracts fingerprint spectrum visual pattern features from the preprocessed protein fingerprint spectrum to obtain the visual pattern feature encoding vector of the fingerprint spectrum of the protein to be tested. These fingerprint spectrum visual pattern features are considered to contain rich macroscopic structural and key region distribution information, similar to an image, and are more resistant to noise and minor changes, thus more stably representing the bacterial species identity. Therefore, by extracting fingerprint spectrum visual pattern features, features that capture deep structural information such as overall morphology, peak distribution patterns, and inter-peak relationships can be extracted from the preprocessed protein fingerprint spectrum.

[0034] Specifically, in one embodiment, extracting visual pattern features from the preprocessed protein fingerprint spectrum to obtain the visual pattern features of the protein fingerprint spectrum to be tested includes: passing the preprocessed protein fingerprint spectrum through a fingerprint spectrum visual pattern feature extractor based on a dilated convolutional neural network model to obtain a visual pattern feature encoding vector for the protein fingerprint spectrum to be tested, which serves as the visual pattern features of the protein fingerprint spectrum to be tested. Models like dilated convolutional neural networks are used to leverage their ability to effectively expand the receptive field, capture long-range dependencies, and multi-scale features when processing image-like data of fingerprint spectra, thereby more comprehensively capturing complex global and local pattern features in the mass spectrum, rather than focusing solely on isolated peaks. The ultimate goal is to transform these complex visual pattern features into a compact and information-rich mathematical representation, namely, the visual pattern feature encoding vector of the protein fingerprint spectrum to be tested. This encoding vector is a high-level generalization and abstraction of the original spectral visual features, aiming to retain discriminative information to the maximum extent while filtering out irrelevant variations.

[0035] Specifically, dilated convolution introduces an interval sampling mechanism, which can significantly expand the receptive field and take into account both local peaks and global trends while keeping the kernel parameters constant. The network structure is generally set as a multi-layer stack, with each layer using a different dilation rate to capture multi-scale feature information. In a specific embodiment, the fingerprint spectrum visual pattern feature extractor based on the dilated convolutional neural network model first sets three convolutional layers sequentially: the first convolutional layer (64 channels, kernel width 3, dilation rate 1), the second convolutional layer (128 channels, kernel width 3, dilation rate 2), and the third convolutional layer (256 channels, kernel width 3, dilation rate 4). Batch normalization and ReLU activation are inserted between each layer, and global average pooling is added as the output layer. The spatial feature vectors of all channels are averaged along the channel direction and compressed into a 256-dimensional visual pattern feature encoding vector for the protein fingerprint spectrum. During the training phase, a large number of bacterial species fingerprint spectra and their category labels from a reference database are required. Supervised learning is used, defining cross-entropy loss or triplet loss functions to guide the classification and segmentation of the feature encoding space. Parameter optimization utilizes adaptive gradient descent methods such as Adam, setting the learning rate to 1e-3 to 1e-5, batch size to 32 to 128, and training epochs to 10 to 100, depending on the training set size and model convergence properties. During model training, mechanisms such as Dropout and L2 regularization can be introduced to prevent overfitting, and data augmentation strategies such as spectral shifting and peak width perturbation can be used to enrich data diversity. Finally, the parameters after training convergence are frozen and used as a fingerprint spectrum visual pattern feature extractor during the inference stage.

[0036] Specifically, in step S5, fingerprint visual pattern features are extracted from the set of protein reference fingerprint spectra of the known bacterial species to obtain a set of protein fingerprint reference visual pattern features. It should be understood that in order to construct a high-quality reference benchmark that matches the feature representation of the sample to be tested, and to overcome the shortcomings of traditional methods that rely on discrete peak information and are easily interfered with, it is necessary not only to extract such robust features from the sample to be tested, but also to perform the same feature transformation processing on the known bacterial species reference database used for comparison. If the reference database still uses the traditional spectra or peak list format, it will be unable to effectively compare and match with the visual pattern feature encoding vector of the sample to be tested, rendering the subsequent query network ineffective. Therefore, in the technical solution of this application, fingerprint visual pattern features are further extracted from the set of protein reference fingerprint spectra of the known bacterial species to obtain a set of protein fingerprint reference visual pattern feature encoding vectors. In other words, the raw or preprocessed protein fingerprint spectra of a large number of known bacterial species (including different Helicobacter pylori strains and other related species) stored in the bacterial species reference database are transformed into a standardized, visual pattern feature-based database using the same feature extraction mechanism (e.g., employing a dilated convolutional neural network model) as the samples being tested. Specifically, a corresponding protein fingerprint reference visual pattern feature encoding vector is generated for each reference bacterial species spectra, and these vectors are collected to form a new, feature domain-based reference database. The goal of this database is to accurately and stably encode the feature identities of each known bacterial species, ensuring that its representation is consistent with the feature vectors of the samples being tested, thus laying the foundation for subsequent pattern matching.

[0037] In one embodiment, extracting fingerprint visual pattern features from the set of protein reference fingerprint spectra of the known bacterial species to obtain a set of protein fingerprint reference visual pattern features includes: passing the protein reference fingerprint spectra of each known bacterial species in the set of protein reference fingerprint spectra of the known bacterial species through the fingerprint visual pattern feature extractor based on the dilated convolutional neural network model to obtain a set of protein fingerprint reference visual pattern feature encoding vectors, which is then used as the set of protein fingerprint reference visual pattern features.

[0038] Specifically, in step S6, a fingerprint pattern feature query is performed on the set of visual pattern features of the protein fingerprint spectrum to be tested and the set of reference visual pattern features of the protein fingerprint spectrum to obtain the fingerprint pattern visual pattern feature query response feature. It should be understood that traditional methods based on peak list comparison or overall spectrum similarity calculation often remain at the level of surface feature similarity comparison, making it difficult to effectively capture complex pattern changes caused by differences in protein expression, sample processing, or instrument fluctuations. These methods are prone to failure when distinguishing closely related species or processing variant spectra. Therefore, simply extracting more robust visual pattern features is insufficient; an intelligent engine capable of understanding and comparing these complex feature patterns is also needed. Therefore, in the technical solution of this application, the set of the encoded vectors of the visual pattern features of the protein fingerprint spectrum to be tested and the encoded vectors of the reference visual pattern features of the protein fingerprint spectrum is further input into a fingerprint pattern feature query network to obtain the fingerprint pattern visual pattern feature query response encoded vector. In particular, the fingerprint pattern feature query network can achieve high-level, in-depth comparisons between the visual pattern features of the sample to be tested and numerous reference pattern features in the reference database, ultimately condensing a discriminative signal that accurately reflects the attribution of the sample to be tested. This fingerprint pattern feature query network can perform single-unit semantic query and self-attention aggregation. First, it uses single-unit semantic query units to meticulously evaluate the semantic relevance between the test feature vector and each reference feature vector. Here, semantics refers to the species-specific biological information contained in the mass spectrometry visual pattern. This is not merely calculating a simple similarity score, but generating a single-unit semantic query score encoding vector for the test protein fingerprint visual pattern that contains multi-dimensional matching information. More importantly, the network then utilizes a self-attention aggregation strategy and considers the self-distribution characteristics of the feature set to achieve context-aware matching evaluation. This means that the network does not view the matching results of each test-reference pair in isolation, but considers the test sample within the context of the entire reference database, dynamically determining which reference patterns have the most significant and consistent global match with the test sample, and adjusting the weights accordingly to effectively aggregate information. This fingerprint spectrum visual pattern feature query response encoding vector realizes the leap from scattered local feature matching to unified, global semantic judgment, providing a solid and highly confident basis for determining the final bacterial species detection result based on this vector, thus strongly supporting the goal of the entire technical solution to achieve more reliable and accurate Helicobacter pylori mass spectrometry identification.

[0039] In one embodiment, as shown in Figure 3, performing fingerprint pattern feature query on the set of visual pattern features of the protein fingerprint spectrum to be tested and the set of reference visual pattern features of the protein fingerprint spectrum to obtain fingerprint pattern visual pattern feature query response features includes: S61, after feature enhancement of the visual pattern feature encoding vector of the protein fingerprint spectrum to be tested, performing individual semantic query encoding on it and each of the reference visual pattern feature encoding vectors of the protein fingerprint spectrum in the set of reference visual pattern feature encoding vectors to obtain a set of individual semantic query score encoding vectors of the visual pattern of the protein fingerprint spectrum to be tested; S62, calculating the individual semantic weight of each individual semantic query score encoding vector of the visual pattern of the protein fingerprint spectrum to be tested in the set of individual semantic query score encoding vectors to obtain a set of individual query semantic self-attention weights of the visual pattern of the protein fingerprint spectrum to be tested; S63, aggregating the set of individual semantic query score encoding vectors of the visual pattern of the protein fingerprint spectrum to be tested based on the set of individual query semantic self-attention weights of the visual pattern of the protein fingerprint spectrum to obtain the fingerprint pattern visual pattern feature query response encoding vector.

[0040] In one embodiment, as shown in Figure 4, after feature enhancement of the visual pattern feature encoding vector of the protein fingerprint spectrum to be tested, it is subjected to individual semantic query encoding with each protein fingerprint spectrum reference visual pattern feature encoding vector in the set of protein fingerprint spectrum reference visual pattern feature encoding vectors to obtain a set of individual semantic query score encoding vectors of the visual pattern of the protein fingerprint spectrum to be tested. This includes: S611, performing feature enhancement based on deconvolution encoding on the visual pattern feature encoding vector of the protein fingerprint spectrum to be tested to obtain an enhanced visual pattern feature encoding vector of the protein fingerprint spectrum to be tested. The enhanced visual pattern feature encoding vector of the protein fingerprint spectrum to be tested has the same feature scale as each protein fingerprint spectrum reference visual pattern feature encoding vector in the set of protein fingerprint spectrum reference visual pattern feature encoding vectors. Specifically, this process can be expressed by the formula:

[0041] ;

[0042] in, The visual pattern feature encoding vector of the protein fingerprint spectrum to be tested. It is deconvolutional encoding. The deconvolution weight matrix is... Let the first norm of the vector be 1. To enhance the visual pattern feature encoding vector of the protein fingerprint spectrum to be tested.

[0043] Specifically, the core effect of deconvolutional coding-based feature enhancement on the visual pattern feature encoding vector of the protein fingerprint spectrum lies in significantly improving the expressive power and semantic richness of the features. Deconvolutional coding (i.e., inverse convolution or transposed convolution) learns data-driven feature upsampling weights, enabling feature reconstruction and enhancement in high-dimensional space while preserving the original key spectral structure information. Specifically, this operation not only improves the spatial resolution of the sample features, strengthening the representation of micro-patterns and global structures, but also adaptively enhances the content of the original encoding vector through a trainable weight system, effectively restoring, supplementing, or highlighting potential but not fully expressed differential and discriminative features in the fingerprint spectrum. Compared to direct linear transformation or ordinary interpolation, the data-adaptive feature enhancement brought about by deconvolutional coding significantly expands the basis for comparing samples with database reference feature vectors, stretching each vector into the same high-level feature space. Furthermore, the deconvolution process ensures complete consistency between the enhancement vector and the feature encoding vectors of all protein fingerprint spectra visual patterns in the reference database in terms of feature scale, dimension, and internal structure distribution. This standardization not only provides a technical guarantee for subsequent fair matching but also significantly improves the generalization and robustness of the pattern recognition system.

[0044] S612, perform individual semantic query encoding on each protein fingerprint reference visual pattern feature encoding vector in the set of enhanced protein fingerprint spectrum visual pattern feature encoding vectors and the set of protein fingerprint spectrum reference visual pattern feature encoding vectors to obtain the set of individual semantic query score encoding vectors of the protein fingerprint spectrum visual pattern. Specifically, this process can be expressed by the formula:

[0045] ;

[0046] ;

[0047] in, This is a set of feature encoding vectors for protein fingerprint spectra referencing visual patterns. These are the 1st, 2nd, and 3rd elements in the set of protein fingerprint reference visual pattern feature encoding vectors. The and the first Each protein fingerprint spectrum references a visual pattern feature encoding vector. and These are the trainable weight matrix and the trainable bias vector, respectively. This is a vector concatenation operation. It is the hyperbolic tangent function. For the first A single semantic query score encoding vector for the visual pattern of the fingerprint spectrum of the protein to be tested.

[0048] Specifically, the set of enhanced visual pattern feature encoding vectors for the protein fingerprint spectrum and the set of reference visual pattern feature encoding vectors for the protein fingerprint spectrum is used to perform individual semantic query encoding to obtain a set of individual semantic query score encoding vectors for the visual pattern of the protein fingerprint spectrum under test. Here, the actual effect of the individual semantic query encoding module is that, through deep feature interaction of the neural network, it automatically learns and quantifies the deep semantic association and accurate matching degree of the corresponding feature pairs formed by the enhanced visual pattern feature encoding vectors for the protein fingerprint spectrum and the reference visual pattern feature encoding vectors for the protein fingerprint spectrum. For example, for different Helicobacter pylori strains or closely related species under the same disease background, although their protein fingerprint visual patterns are similar overall, their deep distribution structure, peak configuration, or sub-block patterns will show rich differences. Individual semantic query can encode this series of semantic information in a high dimension in the matching vector. The individual semantic query score encoding vector for the visual pattern of the protein fingerprint spectrum under test is not a single score, but reflects all the association patterns of the corresponding feature pairs in multiple semantic subspaces, making the pairing between the test spectrum and all database reference spectrums present a rich and detailed dynamic feature relationship distribution.

[0049] In one embodiment, as shown in Figure 5, the individual semantic weights of each individual semantic query score encoding vector in the set of individual semantic query score encoding vectors of the visual pattern of the protein fingerprint spectrum to be tested are calculated to obtain a set of self-attention weights for individual semantic queries of the visual pattern of the protein fingerprint spectrum to be tested. This includes: S621, determining the individual semantic matching degree of each individual semantic query score encoding vector in the set of individual semantic query score encoding vectors of the visual pattern of the protein fingerprint spectrum to be tested based on the feature set self-distribution characteristics of the set of individual semantic query score encoding vectors of the visual pattern of the protein fingerprint spectrum to be tested, to obtain a set of individual semantic matching degrees for the visual pattern of the protein fingerprint spectrum to be tested; S622, inputting the set of individual semantic matching degrees for the visual pattern of the protein fingerprint spectrum to be tested into the relational gating proxy module to obtain a set of self-attention weights for individual semantic queries of the visual pattern of the protein fingerprint spectrum to be tested.

[0050] In one embodiment, in step S621, based on the feature set self-distribution characteristics of the set of visual pattern single semantic query score encoding vectors of the protein fingerprint spectrum to be tested, the single semantic matching degree of each single semantic query score encoding vector of the visual pattern single semantic query score encoding vector of the protein fingerprint spectrum to be tested in the set of visual pattern single semantic query score encoding vectors of the protein fingerprint spectrum to be tested is determined to obtain the set of single semantic matching degrees of the visual pattern single semantic query score encoding vectors of the protein fingerprint spectrum to be tested. This includes: performing inherent alignment calibration on each single semantic query score encoding vector of the visual pattern single semantic query score encoding vector of the protein fingerprint spectrum to be tested in the set of visual pattern single semantic query score encoding vectors of the protein fingerprint spectrum to be tested to obtain the set of calibrated single semantic query score encoding vectors of the visual pattern single semantic query score encoding vectors of the protein fingerprint spectrum to be tested. Specifically, this process can be expressed by the formula:

[0051] ;

[0052] ;

[0053] ;

[0054] ;

[0055] ;

[0056] in, This is the visual pattern query potential vector for the fingerprint spectrum of the protein to be tested. The projection weight matrix is... This is the visual pattern query interaction encoding vector for the protein fingerprint spectrum to be tested. This is a coupling constant, calculated, for example, in the same way as during deconvolution enhancement to maintain symmetry. The visual pattern query covariance matrix for the protein fingerprint spectrum to be tested. For the first A calibrated visual pattern single-unit semantic query score encoding vector for the protein fingerprint spectrum to be tested.

[0057] It can be understood that the single-unit semantic query encoding adopts a method of directly concatenating the enhanced encoding vector of the visual pattern features of the protein fingerprint spectrum to be tested with the corresponding reference visual pattern feature encoding vector of the protein fingerprint spectrum in the database, so as to generate the single-unit semantic query score encoding vector of the visual pattern of the protein fingerprint spectrum to be tested. In this process, it is expected that by leveraging the dynamic characteristics of the representation optimization from the original feature space to the semantic query encoding space, and focusing on calibrating the inherent alignment between the feature space and the semantic query encoding space, the representation accuracy of the single-unit semantic query score encoding vector of the visual pattern of the protein fingerprint spectrum to be tested can be significantly improved.

[0058] Specifically, the enhanced visual pattern feature encoding vector of the protein fingerprint spectrum under test and the reference visual pattern feature encoding vectors of each protein fingerprint spectrum are synthesized into a dynamic eigenvector representation, namely the visual pattern query interaction encoding vector of the protein fingerprint spectrum under test. This visual pattern query interaction encoding vector of the protein fingerprint spectrum under test is further spatially mapped to obtain its projection representation to the semantic query encoding eigenstate, namely the visual pattern query potential vector of the protein fingerprint spectrum under test. On this basis, intrinsic alignment is performed on the visual pattern query potential vector of the protein fingerprint spectrum under test under test under the mapped representation. This intrinsic alignment uses the same coupling constant setting as in the deconvolution feature enhancement step, which effectively ensures the symmetry and mathematical consistency of the entire mapping process. At the same time, by reasonably constructing and solving the covariant matrix of the visual pattern query of the protein fingerprint spectrum under test, the covariant path is constrained under the condition of normal symmetry, thereby fixing the mapping norm. This mechanism improves the representation accuracy of the individual semantic query score encoding vector of the visual pattern of the protein fingerprint spectrum under test in high-dimensional semantic space.

[0059] Based on the distribution characteristics of the set of calibrated protein fingerprint visual pattern monomer semantic query score encoding vectors, the monomer semantic matching degree of each calibrated protein fingerprint visual pattern monomer semantic query score encoding vector is calculated to obtain the set of monomer semantic matching degrees of the protein fingerprint visual pattern. Specifically, this process can be expressed by the formula:

[0060]

[0061] in, for The number of For the first Each calibrated protein fingerprint visual pattern monomer semantic query score encoding vector. Indicated transpose vector, It is a natural exponential function with base e. for function, The semantic matching degree of the visual pattern of the protein fingerprint spectrum to be tested is the monomer semantic matching degree.

[0062] Specifically, the calculation of the individual semantic matching degree is based not only on the numerical characteristics of the calibrated individual semantic query score encoding vector of the visual pattern of the protein fingerprint spectrum, but also on the distribution characteristics of the set of calibrated individual semantic query score encoding vectors of the visual pattern of the protein fingerprint spectrum. In this way, feature pairs that can significantly distinguish the test spectrum from the reference database and have high biological relevance and discriminative power will be assigned higher matching degrees, while noisy, redundant, or weakly discriminative feature pairs will have their influence automatically weakened. Ultimately, all individual semantic matching degrees together constitute the set of individual semantic matching degrees of the visual pattern of the protein fingerprint spectrum.

[0063] Specifically, the calculation process of inputting the set of semantic matching degrees of the visual pattern monomers of the protein fingerprint spectrum to be tested into the relation gating proxy module to obtain the set of semantic self-attention weights of the query semantics of the visual pattern monomers of the protein fingerprint spectrum can be expressed by the following formula:

[0064]

[0065] in, For masking functions, This is a preset threshold, such as 0.5. Of course, this is just an example, and it can be adjusted according to the actual situation. The semantic self-attention weights for the visual pattern query of the protein fingerprint spectrum to be tested are used.

[0066] Specifically, the relation-gated proxy module, leveraging a trainable gating mechanism, effectively preserves the semantic matching scores of individual units in the visual pattern spectrum of the target protein fingerprint spectrum that are most relevant to the discrimination of target bacteria such as Helicobacter pylori. Simultaneously, it automatically suppresses the adverse effects of accidental high scores, weak correlations, or redundant pairings on the overall decision-making result. Based on contextual understanding of the global feature space, the relation-gated proxy module dynamically assigns weights to the semantic matching scores of individual units in the visual pattern spectrum of the target protein fingerprint spectrum during subsequent self-attention aggregation, ensuring that the region focused by the final self-attention mechanism closely matches the actual discrimination requirements.

[0067] Specifically, the calculation process of aggregating the set of semantic query score encoding vectors of the visual pattern of the protein fingerprint spectrum based on the set of semantic self-attention weights of the single query in the fingerprint spectrum to obtain the encoding vector of the fingerprint spectrum visual pattern feature query can be expressed by the following formula:

[0068]

[0069] in, This is the encoding vector for the visual pattern feature query response of the fingerprint spectrum.

[0070] First, through deep modeling of the feature domain, individual semantic query score encoding vectors were obtained for each visual pattern of the protein fingerprint spectrum to be tested and the database reference pattern. Then, utilizing the contextual distribution characteristics of the feature set and a relational gating proxy mechanism, adaptive individual query semantic self-attention weights were assigned to each encoding vector. These weights reflect the actual contribution and discriminative power of each encoding vector in the overall feature representation, effectively highlighting highly discriminative matching information while weakening or suppressing noise and redundancy. Based on this, the aggregation step dynamically and linearly combines all individual semantic query score encoding vectors using the self-attention weights as weighting coefficients. During the aggregation process, not only are local matching information at the micro-level weighted and integrated according to their importance, but a global semantic abstraction and consistent expression of the overall protein fingerprint visual structure map are also achieved.

[0071] Specifically, in step S7, the bacterial species detection result is determined based on the fingerprint spectrum visual pattern feature query response feature. In one embodiment, determining the bacterial species detection result based on the fingerprint spectrum visual pattern feature query response feature includes: passing the fingerprint spectrum visual pattern feature query response encoding vector through a classifier-based bacterial species detector to obtain the bacterial species detection result, which is used to represent the bacterial species category label. Specifically, the Softmax classifier, due to its rigorous structure, strong differentiability, and suitability for multi-class probability modeling, is widely used in the final multi-class discrimination module after deep learning features. In a specific embodiment of this application, the fingerprint spectrum visual pattern feature query response encoding vector is input into the Softmax classifier. If the fingerprint spectrum visual pattern feature query response encoding vector is a fixed-length high-dimensional vector, and there are m bacterial species categories in the system reference database, then for each category i to be discriminated, the Softmax classifier first multiplies it by the learnable weight vector of that category, adds the corresponding bias, obtains the corresponding score, further performs exponential operations to obtain the probability distribution, and finally outputs the category corresponding to the largest pi as the bacterial species detection result, and this probability distribution can be directly regarded as a confidence indicator of the model.

[0072] In summary, the Helicobacter pylori detection method based on mass spectrometry identification provided in this application first acquires the protein fingerprint spectrum of the sample under test using MALDI mass spectrometry. Then, through preprocessing to remove noise and standardization, visual pattern features with discriminative capabilities are extracted. Simultaneously, reference visual pattern features for each known bacterial species are extracted from a pre-constructed bacterial species reference database. Finally, by introducing feature domain-based protein fingerprint spectrum pattern feature query response analysis, the complex matching relationship between the target feature and the reference features is learned to obtain the encoded representation of the fingerprint spectrum visual pattern feature query response. Based on this query response, the bacterial species detection result is finally determined. This strategy based on visual pattern feature extraction and query network matching aims to fully utilize the deep pattern information of mass spectrometry data to achieve more reliable Helicobacter pylori mass spectrometry identification.

[0073] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details of the above embodiments are only for illustrative and facilitative purposes, and are not limitations. These details do not limit the application to the necessity of employing the aforementioned specific details for implementation.

[0074] In the above embodiments, the descriptions of each embodiment have their own emphasis. Parts not described in detail or in a particular embodiment can be referred to in the relevant descriptions of other embodiments. It should be understood that the disclosed devices and methods can be implemented in other ways in the several embodiments provided in this application. For example, the device embodiments described above are merely illustrative; for example, the module division is only a logical functional division, and other division methods may exist in actual implementation. The modules described as separate components may or may not be physically separated. The components shown as modules may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0075] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0076] Finally, it should be noted that the above description has been given for illustrative and descriptive purposes. Furthermore, the above embodiments are only used to illustrate the technical solutions of this application and are not intended to limit it. Although this application has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of this application without departing from the spirit and scope of the technical solutions of this application.

Claims

1. A method for detecting Helicobacter pylori based on mass spectrometry recognition, characterized by, include: The sample to be tested was mixed with the MALDI matrix and spotted onto the MALDI target plate, and the protein fingerprint was acquired using a MALDI-TOF mass spectrometer. Extract a set of protein reference fingerprint profiles for known bacterial species from a bacterial species reference database; The protein fingerprint spectrum is preprocessed to obtain a preprocessed protein fingerprint spectrum; Extracting visual pattern features from the preprocessed protein fingerprint spectrum to obtain visual pattern features of the protein fingerprint spectrum to be tested includes: passing the preprocessed protein fingerprint spectrum through a fingerprint spectrum visual pattern feature extractor based on a dilated convolutional neural network model to obtain a visual pattern feature encoding vector of the protein fingerprint spectrum to be tested as the visual pattern features of the protein fingerprint spectrum to be tested. Extracting fingerprint visual pattern features from the set of protein reference fingerprints of the known bacterial species to obtain a set of protein fingerprint reference visual pattern features includes: passing the protein reference fingerprints of each known bacterial species in the set of protein reference fingerprints of the known bacterial species through the fingerprint visual pattern feature extractor based on the dilated convolutional neural network model to obtain a set of protein fingerprint reference visual pattern feature encoding vectors as the set of protein fingerprint reference visual pattern features; A fingerprint pattern feature query is performed on the set of visual pattern features of the protein fingerprint spectrum to be tested and the set of reference visual pattern features of the protein fingerprint spectrum to obtain the fingerprint pattern visual pattern feature query response features. Based on the visual pattern feature query response features of the fingerprint spectrum, the bacterial species detection result is determined; Specifically, the process involves performing a fingerprint pattern feature query on the set of visual pattern features of the protein fingerprint spectrum to be tested and the set of reference visual pattern features of the protein fingerprint spectrum to obtain fingerprint pattern visual pattern feature query response features, including: After feature enhancement, the visual pattern feature encoding vector of the protein fingerprint spectrum to be tested is subjected to individual semantic query encoding with each protein fingerprint spectrum reference visual pattern feature encoding vector in the set of protein fingerprint spectrum reference visual pattern feature encoding vectors to obtain a set of individual semantic query score encoding vectors of the visual pattern of the protein fingerprint spectrum to be tested. The individual semantic weights of each individual semantic query score encoding vector in the set of individual semantic query score encoding vectors of the visual pattern of the protein fingerprint spectrum are calculated to obtain the set of self-attention weights of the individual semantic query of the visual pattern of the protein fingerprint spectrum. The set of semantic self-attention weights for the visual pattern of the protein fingerprint spectrum under test is used to aggregate the set of semantic query score encoding vectors for the visual pattern of the protein fingerprint spectrum under test to obtain the visual pattern feature query response encoding vector of the fingerprint spectrum.

2. The mass spectrometry-based identification method for detecting H. pylori according to claim 1, characterized in that, The preprocessing includes baseline correction, noise smoothing, peak identification, and alignment.

3. The mass spectrometry-based identification method for detecting H. pylori according to claim 2, characterized in that, After feature enhancement of the visual pattern feature encoding vector of the protein fingerprint spectrum to be tested, it is then subjected to individual semantic query encoding with each protein fingerprint spectrum reference visual pattern feature encoding vector in the set of protein fingerprint spectrum reference visual pattern feature encoding vectors to obtain a set of individual semantic query score encoding vectors of the visual pattern of the protein fingerprint spectrum to be tested, including: The visual pattern feature encoding vector of the protein fingerprint spectrum to be tested is enhanced by deconvolution coding to obtain an enhanced visual pattern feature encoding vector of the protein fingerprint spectrum to be tested. The enhanced visual pattern feature encoding vector of the protein fingerprint spectrum to be tested has the same feature scale as each protein fingerprint spectrum reference visual pattern feature encoding vector in the set of protein fingerprint spectrum reference visual pattern feature encoding vectors. Each protein fingerprint reference visual pattern feature encoding vector in the set of enhanced protein fingerprint spectrum visual pattern feature encoding vectors and the set of protein fingerprint spectrum reference visual pattern feature encoding vectors is subjected to individual semantic query encoding to obtain the set of individual semantic query score encoding vectors of the protein fingerprint spectrum visual pattern.

4. The mass spectrometry-based identification method for detecting H. pylori according to claim 3, characterized in that, The individual semantic weights of each individual semantic query score encoding vector in the set of visual pattern individual semantic query score encoding vectors of the protein fingerprint spectrum are calculated to obtain the set of self-attention weights for individual query semantics of the visual pattern of the protein fingerprint spectrum, including: Based on the feature set self-distribution characteristics of the set of visual pattern single semantic query score encoding vectors of the protein fingerprint spectrum to be tested, the single semantic matching degree of each single semantic query score encoding vector of the visual pattern single semantic query score encoding vector of the protein fingerprint spectrum to be tested is determined in the set of visual pattern single semantic matching degrees of the protein fingerprint spectrum to be tested. The set of semantic matching degrees of the visual pattern monomers of the protein fingerprint spectrum to be tested is input into the relational gating agent module to obtain the set of semantic self-attention weights of the query semantics of the visual pattern monomers of the protein fingerprint spectrum to be tested.

5. The mass spectrometry-based identification method for detecting H. pylori according to claim 4, characterized in that, Based on the self-distribution characteristics of the feature set of the set of semantic query score encoding vectors of the visual pattern of the protein fingerprint spectrum to be tested, the semantic matching degree of each semantic query score encoding vector of the visual pattern of the protein fingerprint spectrum to be tested in the set of semantic query score encoding vectors of the visual pattern of the protein fingerprint spectrum to be tested is determined to obtain the set of semantic matching degrees of the semantic matching degree of the semantic matching degree of the visual pattern of the protein fingerprint spectrum to be tested, including: The inherent alignment of each individual semantic query score encoding vector of the visual pattern of the protein fingerprint spectrum in the set of the semantic query score encoding vectors of the individual molecules of the protein fingerprint spectrum is calibrated to obtain the set of calibrated semantic query score encoding vectors of the individual molecules of the visual pattern of the protein fingerprint spectrum. calculating a monomer semantic matching degree of each of the post-calibration to-be-tested protein fingerprint visual pattern monomer semantic query score encoding vectors based on distribution characteristics of the set of post-calibration to-be-tested protein fingerprint visual pattern monomer semantic query score encoding vectors to obtain a set of the to-be-tested protein fingerprint visual pattern monomer semantic matching degrees.

6. The mass spectrometry-based identification method for detecting H. pylori according to claim 5, characterized in that, based on the fingerprint visual pattern feature query response feature, determining a strain detection result, including: passing the fingerprint visual pattern feature query response encoding vector through a classifier-based strain detector to obtain a strain detection result, the strain detection result being used to represent a strain category label.