Gastroesophageal reflux disease risk prediction esophageal flora image intelligent detection method

Abstract

The present application belongs to the technical field of medical image detection, and particularly relates to an esophageal flora image intelligent detection method for gastroesophageal reflux disease risk prediction, wherein flora image and clinical data are acquired to construct a perception network; flora displacement features are quantified by image registration; flora density, aggregation degree and fluorescence gradient features are extracted by improved U-Net++ segmentation; flora and epithelial cell lesion features are fused by space-time attention to generate flora-host correlation atlas; flora dynamic evolution model is constructed by space-time graph convolution network to predict Barrett esophagus occurrence probability and flora diffusion trend; lesion risk grade is calculated by flora imbalance quantification algorithm; in combination with reflux frequency and pH fluctuation, flora-environment interaction model is used to simulate colonization coefficient and correct parameters, which are transmitted to a diagnosis platform to deduce cancer risk, trigger grading suggestions and generate a visual report. Thus, the problems of long detection cycle and weak microscopic analysis capability in the prior art are solved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of medical image detection, and particularly relates to an esophageal flora image intelligent detection method for gastroesophageal reflux disease risk prediction. BACKGROUND

[0002] With the development of precise medical technology, the market urgently needs an intelligent detection system for reflux esophagitis based on flora-mucosa correlation analysis, which requires multi-modal data fusion, dynamic feature extraction, and multi-dimensional pathological deduction capabilities. However, the existing technology cannot realize deep fusion of flora and mucosal lesion data, lacks flora-host collaborative reasoning framework and dynamic risk quantification mechanism, and is difficult to support precise detection in complex digestive tract scenarios. There is a significant gap between the existing technology and the technical requirements of constructing flora-host correlation atlas and real-time correction of detection sensitivity.

[0003] However, the traditional flora detection method has the following key defects: the detection method relying on gastroscopy biopsy and culture has a long cycle (24-72 hours), and is insufficient in capturing early microflora changes; image analysis can only identify macroscopic lesions such as esophageal mucosal erosion, and lacks micro-analysis of flora community structure and dominant bacterial species morphology; the data dimension is limited to single pathological images or biochemical indicators, and a multi-modal perception network has not been formed, lacking a spatiotemporal attention mechanism for collaborative processing of flora dynamics and epithelial cell lesions, resulting in a data gap between flora characteristics and mucosal damage. With the advancement of precision medicine, the market urgently needs high-sensitivity, low-latency flora intelligent detection technology. However, the existing technology has limited feature expression capability and insufficient detection robustness, making it difficult to support early warning applications in complex scenarios. SUMMARY

[0004] The present application provides an esophageal flora image intelligent detection method for gastroesophageal reflux disease risk prediction to solve the problems of long detection cycle and weak micro-analysis capability in the prior art.

[0005] The first aspect embodiment of the application provides an esophageal flora image intelligent detection method for predicting the risk of gastroesophageal reflux disease, comprising the following steps: obtaining flora microscopic images and clinical feature data of esophageal mucosa samples; constructing a perception network according to the flora microscopic images and the clinical feature data, quantifying the spatial displacement characteristics of the flora sample sequence based on the perception network by using an image registration algorithm, extracting the flora density, aggregation degree and fluorescence intensity gradient characteristics by using an improved U-Net++ network segmentation image, fusing the flora dynamics and epithelial cell lesion characteristics through a spatio-temporal attention mechanism to generate a flora-host correlation atlas; constructing a flora dynamic evolution model by using a spatio-temporal graph convolution network according to the flora-host correlation atlas, predicting the Barrett esophagus occurrence probability and flora diffusion trend, calculating the lesion risk level by using a flora imbalance quantification algorithm based on the predicted Barrett esophagus occurrence probability and flora diffusion trend, combining the patient's gastric acid reflux frequency and esophageal pH value fluctuation data, simulating the flora colonization coefficient by using a flora-environment interaction model, real-time correcting the sensitivity parameters, transmitting the corrected sensitivity parameters to a clinical diagnosis platform, performing multi-dimensional cancer risk deduction, triggering a hierarchical diagnosis suggestion and generating a visual report of a risk heat map.

[0006] Preferably, the visual report of the risk heat map comprises: constructing a flora-cancer collaborative diffusion model; simulating the invasion trajectory of pathogenic bacteria in the esophageal mucosa layer by combining the improved particle swarm algorithm based on the flora-cancer collaborative diffusion model, mapping the spatial correlation between the flora aggregation area and the epithelial cell dysplasia area to the cancer probability value by using a Gaussian mixture model, generating a cancer risk probability cloud map; superimposing the cancer risk probability cloud map on the esophageal three-dimensional anatomical model, representing the cancer risk intensity by color gradient, and generating a heat map visualization report labeling the flora pathogenic key path nodes, the predicted cancer progression time and the high-risk area coordinates.

[0007] Preferably, the spatio-temporal attention mechanism formula is:

[0008]

[0009] wherein is the attention output result; is the flora dynamic feature matrix; is the epithelial cell lesion feature matrix; is the query vector generated by M; is the key vector generated by E; is the transpose matrix of the key vector ; is the key vector dimension; is the value vector; is the scaled flora-host feature correlation score. ​

[0010] Preferably, a microbiota-environment interaction model is used to simulate the microbiota colonization coefficient and correct sensitivity parameters in real time, including: constructing a microbiota-environment interaction dynamic model; inputting data on gastric acid reflux frequency, esophageal pH fluctuations, and microbiota metabolite concentrations into the microbiota-environment interaction dynamic model, simulating the change curve of the microbiota colonization coefficient with environmental factors through differential equations; and correcting sensitivity parameters in real time based on the change curve to control the error between predicted values ​​and clinically measured values.

[0011] Preferably, a spatiotemporal graph convolutional network is used to construct a microbial community dynamic evolution model to obtain the probability of Barrett's esophagus and the microbial community diffusion trend. This includes: constructing a microbial community dynamic evolution model; based on the microbial community dynamic evolution model, setting the high-incidence areas of precancerous lesions in the dentate line of the lower esophagus and the cardia mucosal folds as core nodes of the graph network, and combining the microscopic characteristics of microbial biofilm thickness, epithelial cell apoptosis rate, and inflammatory factor concentration, modeling the spatiotemporal correlation between microbial invasion depth and the degree of epithelial dysplasia through a spatiotemporal graph convolutional network to predict the probability of Barrett's esophagus and the corresponding microbial community diffusion trend.

[0012] Preferably, a perceptual network is constructed, and based on the perceptual network, an image registration algorithm is used to quantify the spatial displacement features of the microbial community in the esophageal sample sequence, including: constructing a multimodal perceptual network; according to the multimodal perceptual network, fusing the texture features of the microscopic image of the microbial community with the structured information of clinical feature data to generate a dynamic weight matrix; based on the dynamic weight matrix, using an image registration algorithm, calculating the three-dimensional spatial displacement vector of the microbial community in the esophageal sample sequence, and quantifying the migration distance and orientation angle parameters of the microbial community.

[0013] Preferably, an improved U-Net++ network is used to segment the image and extract the bacterial community density, aggregation degree, and fluorescence intensity gradient features, including: constructing an improved U-Net++ network; performing pixel-level segmentation on the bacterial community microscopic image based on the improved U-Net++ network to generate a bacterial community region mask; calculating the bacterial community density index using a density clustering algorithm based on the bacterial community region mask, extracting aggregation degree features through morphological operators, and calculating the gradient magnitude and direction features of fluorescence intensity using the Sobel operator.

[0014] Preferably, the improved U-Net++ network formula is as follows:

[0015]

[0016] in, Let be the image segmentation accuracy at time t; Based on the accuracy of segmentation; The fluorescence contrast of the bacterial community at time t; Standard contrast threshold; This represents the regional weighting coefficient of the microbial community. Noise impact factor; Let be the image noise level at time t.

[0017] Preferably, the triggering of graded diagnostic recommendations includes: setting up a four-level diagnostic mechanism. When the probability of Barrett's esophagus is less than 20%, with no abnormal spread of the flora and normal epithelial cells, it is classified as Level 1 low risk, and dietary intervention and follow-up every 6 months are recommended; when the probability of occurrence is 20%-50%, with localized flora aggregation and mild epithelial hyperplasia, it is classified as Level 2 low-to-medium risk, and probiotic intervention and gastroscopy follow-up are recommended; when the probability of occurrence is 50%-80%, with flora spreading to the upper esophagus and accompanied by moderate dysplasia, it is classified as Level 3 medium-to-high risk, triggering evaluation for endoscopic mucosal resection and targeted antimicrobial therapy; when the probability of occurrence is greater than 80%, with flora invasion of the entire esophagus and accompanied by severe dysplasia or early cancerous changes, it is classified as Level 4 high risk, initiating multidisciplinary consultation and evaluation for radical surgery.

[0018] The second aspect of this application provides an intelligent esophageal microbiota image detection system for predicting the risk of gastroesophageal reflux disease, comprising: an acquisition module for acquiring microscopic images of microbiota and clinical feature data of esophageal mucosal samples; and an extraction module for constructing a perception network based on the microscopic images and clinical feature data, quantifying the spatial displacement features of microbiota in esophageal sample sequences using an image registration algorithm based on the perception network, segmenting the image using an improved U-Net++ network, extracting microbiota density, aggregation degree, and fluorescence intensity gradient features, and fusing microbiota dynamics and epithelial cell lesion features through a spatiotemporal attention mechanism to generate a microbiota-host association map. The generation module is used to construct a dynamic evolution model of the microbiome using a spatiotemporal graph convolutional network based on the microbiome-host association map, predict the probability of Barrett's esophagus and the microbiome spread trend, calculate the lesion risk level based on the predicted probability of Barrett's esophagus and the microbiome spread trend through a microbiome dystrophic algorithm, combine the patient's gastric acid reflux frequency and esophageal pH fluctuation data, use a microbiome-environment interaction model to simulate the microbiome colonization coefficient, correct the sensitivity parameters in real time, and transmit the corrected sensitivity parameters to the clinical diagnostic platform to perform multi-dimensional cancer risk extrapolation, trigger graded diagnosis suggestions, and generate a visual report of the risk heatmap.

[0019] Therefore, this application includes the following beneficial effects: The embodiments of this application integrate esophageal mucosal microscopic images with multi-dimensional clinical data, accurately quantify the spatial displacement characteristics of the microbiota using image registration algorithms, and achieve efficient extraction of microscopic features of the microbiota using an improved U-Net++ network. Furthermore, through a spatiotemporal attention mechanism, the dynamic changes of the microbiota and epithelial cell lesion characteristics are deeply fused, generating a microbiota-host association map that breaks the separation between microbiota data and mucosal damage information in traditional detection. This achieves cross-scale association analysis from microscopic microbiota behavior to macroscopic lesions, significantly improving the ability to capture early microbial abnormalities and occult mucosal damage. The microbiota dynamic evolution model constructed based on a spatiotemporal graph convolutional network can accurately predict the probability of Barrett's esophagus and the trend of microbiota diffusion. Combined with environmental data such as the frequency of gastric acid reflux and esophageal pH fluctuations, the detection sensitivity parameters are corrected in real time through a microbiota-environment interaction model, effectively solving the problem of poor adaptability to individual differences and environmental fluctuations in traditional methods. This makes risk assessment more consistent with the patient's actual pathological state, significantly improving the accuracy and timeliness of precancerous lesion warning. The resulting tiered diagnostic recommendations and risk heatmap visualization report provide clinicians with intuitive decision-making support, from the pathogenic pathways of gut microbiota to the risk of cancer. This not only shortens the testing cycle (by more than 60% compared to traditional culture methods) but also reduces the risk of overtreatment and missed diagnoses through precise tiered treatment recommendations. It comprehensively improves the accuracy, clinical suitability, and diagnostic efficiency of reflux esophagitis and related precancerous lesions, providing new technical support for the early intervention and precise management of digestive tract diseases. This solves the problems of long testing cycles and weak microscopic resolution in existing technologies.

[0020] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0021] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0022] Figure 1 This is a flowchart of an intelligent detection method for esophageal microbiota images for predicting the risk of gastroesophageal reflux disease, provided according to an embodiment of this application.

[0023] Figure 2 This is a schematic diagram of microscopic image sequence analysis of gastric mucosal flora according to an embodiment of this application;

[0024] Figure 3 This is a schematic diagram of intelligent detection of bacterial flora in reflux esophagitis according to an embodiment of this application;

[0025] Figure 4This is a schematic diagram of a Barrett's esophageal risk assessment provided according to an embodiment of this application;

[0026] Figure 5 This is a schematic diagram of an intelligent detection method for esophageal microbiota images for predicting the risk of gastroesophageal reflux disease according to an embodiment of this application;

[0027] Figure 6 This is a schematic diagram of the structure of an intelligent esophageal microbiota image detection system for predicting the risk of gastroesophageal reflux disease according to an embodiment of this application;

[0028] Figure 7 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of this application. Detailed Implementation

[0029] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0030] The following describes, with reference to the accompanying drawings, an intelligent detection method for esophageal microbiota images for predicting the risk of gastroesophageal reflux disease (GERD) based on embodiments of this application. Addressing the issue of long detection cycles mentioned in the background section, this application provides an intelligent detection method for esophageal microbiota images for predicting the risk of GERD. This method integrates esophageal mucosal microscopic images with multi-dimensional clinical data, accurately quantifies the spatial displacement characteristics of the microbiota using image registration algorithms, efficiently extracts microscopic features of the microbiota using an improved U-Net++ network, and deeply fuses dynamic changes in the microbiota with epithelial cell lesion features through a spatiotemporal attention mechanism. The generated microbiota-host association map breaks down the separation between microbiota data and mucosal damage information in traditional detection methods, achieving cross-scale association analysis from microscopic microbiota behavior to macroscopic lesions, significantly improving the ability to capture early microbial abnormalities and occult mucosal damage. A microbial dynamic evolution model constructed based on a spatiotemporal graph convolutional network can accurately predict the probability of Barrett's esophagus and the trend of microbial diffusion. Combined with environmental data such as the frequency of gastric acid reflux and esophageal pH fluctuations, the model uses a microbial-environment interaction model to correct detection sensitivity parameters in real time. This effectively solves the problem of poor adaptability to individual differences and environmental fluctuations in traditional methods, making risk assessment more closely aligned with the patient's actual pathological state and significantly improving the accuracy and timeliness of precancerous lesion early warning. The final generated tiered diagnostic recommendations and risk heatmap visualization report provide clinicians with intuitive decision-making basis from the pathogenesis pathway of microbial flora to the risk of cancer. This not only shortens the detection cycle (by more than 60% compared to traditional culture methods) but also reduces the risk of overtreatment and missed diagnosis through precise stratified treatment recommendations. It comprehensively improves the accuracy, clinical suitability, and diagnostic efficiency of reflux esophagitis and related precancerous lesion detection, providing new technical support for the early intervention and precise management of digestive tract diseases. Thus, it solves the problems of long detection cycles and weak microscopic analysis capabilities in existing technologies.

[0031] Specifically, Figure 1 This is a flowchart of an intelligent detection method for esophageal microbiota images for predicting the risk of gastroesophageal reflux disease, provided in an embodiment of this application.

[0032] like Figure 1 As shown, this intelligent detection method for esophageal microbiota images for predicting the risk of gastroesophageal reflux disease includes the following steps:

[0033] In step S101, microscopic images of the microbial community and clinical characteristic data of the esophageal mucosa sample are obtained.

[0034] Clinical characteristic data refers to the collective term for various clinical information related to a patient's disease, such as symptoms, medical history, examination results, and treatment records.

[0035] It is understood that the embodiments of this application can provide microscopic features such as morphology and density of the microbial community by acquiring microscopic images of the microbial community of esophageal mucosa samples. Clinical feature data can reflect macroscopic information such as patient symptoms, medical history and examination results, providing a comprehensive data foundation for constructing a microbial community-host association map and analyzing the association between microbial community and lesions. It can accurately capture microscopic microbial community abnormalities and incorporate the individual clinical background of patients, thereby improving the comprehensiveness and specificity of the detection.

[0036] In step S102, a perceptual network is constructed based on the microscopic images of the microbial community and clinical feature data. Based on the perceptual network, an image registration algorithm is used to quantify the spatial displacement features of the microbial community in the esophageal sample sequence. At the same time, an improved U-Net++ network is used to segment the image and extract the microbial community density, aggregation degree and fluorescence intensity gradient features. The microbial community dynamics and epithelial cell lesion features are fused through a spatiotemporal attention mechanism to generate a microbial community-host association map.

[0037] Image registration algorithms are a technique that uses spatial transformation relationships to spatially align multiple images acquired under different conditions (such as sample images from different times and perspectives) to eliminate positional differences and achieve accurate comparative analysis.

[0038] It is understood that the embodiments of this application, by employing image registration algorithms, can spatially align sample images from different times and perspectives, eliminate positional differences to achieve accurate comparison, accurately quantify the spatial displacement characteristics of bacterial communities in sequence samples, avoid errors in bacterial community dynamic analysis caused by image misalignment, provide a reliable spatial benchmark for extracting features such as bacterial community density and aggregation, more accurately capture dynamic patterns such as bacterial community migration and diffusion, lay a solid data foundation for generating bacterial community-host association maps, and improve the accuracy and reliability of bacterial community dynamic analysis in overall detection.

[0039] It should be noted that the image registration algorithm formula is as follows:

[0040]

[0041] in, The x-coordinate of the transformed point; The ordinate of the transformed point; These are the homogeneous coordinates of the transformed points; These are the homogeneous coordinates of the point before the transformation; The x-coordinate of the point before the transformation; The ordinate of the point before the transformation; This is the core parameter matrix of the affine transformation; This represents the translation in the x-direction; This represents the translation in the y-direction; , , , This is the linear transformation part.

[0042] Microbiome-host association maps are visualization and analysis tools used to systematically analyze the interaction between microbial communities and hosts. Their core function is to reveal the dynamic relationship between the two in physiological metabolism, immune regulation, and disease development through multi-dimensional data association.

[0043] For example, such as Figure 2 As shown, in the microscopic image sequence analysis of gastric mucosal microbiota, when using the image registration algorithm, the first frame of the sequence is used as a reference template to extract the edge contour feature points and specific fluorescent marker anchor points of the microbiota. A scale-invariant feature transformation algorithm is then used to match the homologous features of subsequent frames. To address spatial misalignment caused by sample slice curling and field-of-view shift, the optimal rigid transformation matrix (including rotation angle and translation) is calculated to calibrate each frame to a unified coordinate system. After registration, the spatial coordinate changes of the labeled microbiota are tracked, and the displacement parameters after removing noise interference are quantified. For example, a certain streptococcal community has an average displacement of 15.6 μm over 3 hours, with the displacement direction matching the epithelial cell arrangement direction by 82%. The population displacement dispersion has decreased from 9.3 μm before registration to 3.5 μm. This approach accurately removes spurious displacements caused by physical disturbances in the samples, providing a standardized spatial benchmark for subsequent temporal comparison of microbiota density and spatial distribution analysis of aggregation, ensuring the accuracy of the fusion of microbiota dynamic characteristics and epithelial lesion characteristics, and supporting the quantification of spatial interactions in the microbiota-host association map.

[0044] In this embodiment of the application, the spatiotemporal attention mechanism formula is as follows:

[0045]

[0046] in, Output results for attention; This is the microbial community dynamics feature matrix; A matrix representing the characteristics of epithelial cell lesions; The query vector generated by M; The key vector generated by E; Key vector The transpose of the matrix; The dimension of the key vector; It is a value vector; The scaled score represents the microbiome-host feature association score.

[0047] It is understood that, in the spatiotemporal attention mechanism of this application embodiment, when fusing microbial dynamics and epithelial cell lesion features, the spatial dimension focuses on key areas closely related to microbial aggregation and lesions, weakening irrelevant background interference; the temporal dimension captures the temporal correlation between microbial displacement rhythm and lesion development, prioritizing the enhancement of clinically significant dynamic features, improving the targeting and accuracy of feature fusion, reducing the interference of noise information on association analysis, and making the generated microbial-host association map more clearly reflect the core laws, providing visualized data for clinical interpretation of the causal relationship between microbial abnormalities and disease progression, and enhancing the model's adaptability to complex pathological scenarios.

[0048] For example, when processing esophageal sample sequences, the spatiotemporal attention mechanism first focuses on the time dimension. By analyzing the rate and direction changes of the spatial displacement of the microbial community, it assigns higher attention weights to periods where the displacement amplitude exceeds a threshold, thus highlighting key time points in microbial migration, such as moments when the microbial community suddenly aggregates or spreads in the lesion area. Then, in the spatial dimension, it compares the high-density areas of the microbial community and the bands of dramatic fluorescence intensity segmented by the improved U-Net++ network with the morphological features of the epithelial cell lesion area, giving greater attention to areas with high overlap and dynamically adjusting the spatial weight values. Through this synergistic optimization of spatiotemporal weights, irrelevant noise information can be effectively filtered out, significantly strengthening the core correlation information between microbial community dynamics and lesion features during the fusion process. This allows for a precise delineation of the spatiotemporal linkage patterns between the two in different time periods and regions, providing more accurate feature support for the construction of microbial-host association maps.

[0049] It should be noted that the displacement amplitude threshold is not a fixed value and needs to be dynamically set according to the specific scenario. It usually depends on the sample type (such as intestinal or skin samples), the type of microbial community (such as the differences in the motility characteristics of cocci and bacilli), and the imaging resolution (such as the actual spatial scale corresponding to micron-level pixels). For cocci communities in intestinal mucosal samples, if the imaging resolution is 0.5 μm / pixel, the threshold is set to 5-10 pixels (i.e., 2.5-5 μm), because this range can effectively distinguish between random displacement of the microbial community caused by Brownian motion and directional migration driven by the pathological microenvironment; while for more motile bacilli, the threshold is increased to 15-20 pixels (7.5-10 μm) to capture their significant aggregation or diffusion behavior.

[0050] In this embodiment, a perceptual network is constructed, and based on the perceptual network, an image registration algorithm is used to quantify the spatial displacement characteristics of the microbial community in the esophageal sample sequence. This includes: constructing a multimodal perceptual network; based on the multimodal perceptual network, fusing the texture features of the microscopic image of the microbial community with the structured information of clinical feature data to generate a dynamic weight matrix; and based on the dynamic weight matrix, using an image registration algorithm to calculate the three-dimensional spatial displacement vector of the microbial community in the esophageal sample sequence, and quantifying the migration distance and orientation angle parameters of the microbial community.

[0051] Among them, the multimodal perception network is a network model that integrates multiple modal data such as microscopic images of microbial communities and clinical features, and extracts key cross-modal information through collaborative perception, providing multi-source data integration support for microbial community-host association analysis.

[0052] It is understood that the embodiments of this application utilize a multimodal sensing network to effectively integrate the texture features of microscopic images of microbiota with the structured information of clinical feature data. Through cross-modal collaborative sensing, more comprehensive key information is extracted, and the generated dynamic weight matrix can accurately provide data for image registration algorithms. This makes the calculation of the three-dimensional spatial displacement vector of microbiota in esophageal sample sequences, the quantification of migration distance and orientation angle parameters more in line with clinical practice, improves the completeness of feature extraction, and enhances the accuracy of subsequent quantitative analysis.

[0053] For example, such as Figure 3 As shown, in the intelligent detection of esophageal microbiota images in reflux esophagitis, the multimodal perception network deeply integrates multidimensional texture features of esophageal microscopic images (such as gradient changes in microbiota aggregation density, differences in attachment morphology to damaged epithelium, and intensity distribution of fluorescently labeled microbiota) with multidimensional clinical feature data (such as acid exposure time from 24-hour esophageal pH monitoring, endoscopic inflammation scores, patient reflux frequency, and symptom duration). Through cross-modal association mining, it accurately captures the potential correlation between microbiota colonization patterns and disease progression, generating a dynamic weight matrix—for example, assigning higher weights to microbiota adhesion texture in erosion areas and the proportion of total time with pH < 4 during the acute phase, while strengthening the weights of microbiota biofilm formation texture and tissue fibrosis degree during the chronic phase. Based on this matrix, combined with image registration algorithms, the three-dimensional spatial displacement of the microbiota in the inflammatory area (such as the migration distance to the erosion site and the diffusion angle along the longitudinal axis of the esophagus) can be accurately quantified, enabling precise identification of the abnormal colonization patterns of the characteristic microbiota in reflux esophagitis (such as the abnormal adhesion of Helicobacter pylori to the squamous epithelium), improving the detection specificity by 15%-20% and greatly enhancing the fit with clinical diagnosis and treatment needs.

[0054] In this embodiment, an improved U-Net++ network is used to segment images and extract bacterial community density, aggregation degree, and fluorescence intensity gradient features. This includes: constructing an improved U-Net++ network; performing pixel-level segmentation on the bacterial community microscopic image based on the improved U-Net++ network to generate a bacterial community region mask; calculating the bacterial community density index using a density clustering algorithm based on the bacterial community region mask; extracting aggregation degree features using morphological operators; and calculating the gradient magnitude and direction features of fluorescence intensity using the Sobe operator.

[0055] Density clustering algorithm is a method of clustering based on the spatial density of sample points. It divides samples with connected densities into the same cluster and identifies noise points in low-density areas, thereby achieving cluster analysis of data with complex distribution patterns.

[0056] It is understood that the embodiments of this application employ density clustering algorithms to accurately process the complex patterns of non-uniform bacterial distribution based on bacterial community region masks. By identifying low-density noise points to remove interfering information, and dividing densely connected bacterial communities into clusters to calculate density indices, the accuracy of bacterial community density quantification is improved. This provides a reliable basis for subsequent extraction of aggregation features using morphological operators and calculation of fluorescence intensity gradient features using the Sobel operator, accurately capturing the distribution characteristics of bacterial communities in lesion areas and enhancing the effectiveness of bacterial community-lesion correlation analysis.

[0057] It should be noted that the microbial community region mask is a binary image generated after pixel-level segmentation of the microbial community image. The region where the microbial community is located is marked (e.g., the pixel value is 1), while the background region is marked with other values ​​(e.g., the pixel value is 0), which is used to accurately define the spatial range of the microbial community.

[0058] Density clustering algorithm formula:

[0059]

[0060] in, The data points to be determined; With p as the center and radius as The set of all data points contained within the area; The minimum number of neighborhood points specified by the user.

[0061] Morphological operator formulas:

[0062]

[0063]

[0064]

[0065]

[0066] in, This is the set of pixels corresponding to the mask of the bacterial community region; For structural elements; The coordinate vector of the pixel; These are the relative coordinates of the pixels within structuring element B.

[0067] Sobe operator formula:

[0068]

[0069]

[0070]

[0071]

[0072] in, Microscopic images of bacterial communities; This represents the gradient response of the image in the x-direction; This represents the gradient response of the image in the y-direction. This represents the gradient magnitude. The gradient direction.

[0073] For example, in microscopic image analysis of bacterial communities, density clustering algorithms (such as DBSCAN) can accurately identify bacterial community aggregation regions. For instance, for a bacterial community region mask generated by segmentation using an improved U-Net++ network, a neighborhood radius ε (e.g., 3 pixels) and a minimum number of points MinPts (e.g., 5 pixels) are first set, and the number of pixels within the ε-neighborhood of each bacterial community pixel is calculated. If the number of pixels within the neighborhood of a pixel is greater than or equal to MinPts, it is identified as a core point. Pixels within the ε-neighborhood of the core point form "clusters" (dense bacterial communities) based on density reachability, while pixels not covered by any core point are marked as noise points (scattered bacterial communities). By statistically analyzing the ratio of the total number of pixels within a cluster to the cluster area, the bacterial community density index can be quantified, aiding in the analysis of bacterial colonization patterns in scenarios such as reflux esophagitis. This method does not require a preset number of clusters, can automatically discover aggregation regions of arbitrary shapes, and is robust to noise, significantly outperforming traditional algorithms such as K-means.

[0074] In this embodiment, the improved U-Net++ network formula is as follows:

[0075]

[0076] in, Let be the image segmentation accuracy at time t; Based on the accuracy of segmentation; The fluorescence contrast of the bacterial community at time t; Standard contrast threshold; This represents the regional weighting coefficient of the microbial community. Noise impact factor; Let be the image noise level at time t.

[0077] It is understood that the embodiments of this application can accurately segment microscopic images of bacterial communities by using an improved U-Net++ network, effectively extracting key features such as bacterial community density, aggregation degree, and fluorescence intensity gradient. The optimized network structure enhances the multi-scale feature fusion capability, keenly captures subtle morphological and spatial distribution differences of bacterial communities, and improves segmentation accuracy and the completeness of feature extraction. At the same time, for scenarios where bacterial communities and epithelial cells are complexly intertwined in samples, the improved network has a stronger ability to integrate contextual information, providing high-quality basic data support for subsequent spatiotemporal attention mechanism fusion of bacterial community dynamics and lesion features to generate accurate bacterial community-host association maps, thereby improving the reliability of clinical feature analysis.

[0078] For example, in the analysis of fluorescence microscopic images of gut microbiota samples, the improved U-Net++ network significantly improved the segmentation accuracy of dense microbial communities by introducing a multi-scale feature fusion module and a spatial attention mechanism. For instance, when processing images with weak fluorescence signals and complex interweaving of microbiota and epithelial cells, the network dynamically fuses feature information from different levels through an optimized dense skip connection structure. This not only accurately segments the boundaries of the microbiota (Dice coefficient exceeding 0.95) but also precisely extracts the microbiota density gradient, aggregation morphology, and fluorescence intensity distribution features. Experiments show that the improved network achieves an accuracy of 95% in images containing more than 10,000 bacterial cells and can effectively distinguish complex clumps formed by microbial aggregation from surrounding host tissue. Its high-resolution segmentation output provides pixel-level localization evidence for subsequent quantification of the spatial displacement of microbiota relative to diseased epithelial cells (such as detecting microbial migration paths through image registration algorithms), and lays the foundation for generating microbiota-host association maps, helping to reveal the potential correlation between microbiota dynamics and host lesions.

[0079] In step S103, based on the microbiota-host association map, a dynamic evolution model of the microbiota is constructed using a spatiotemporal graph convolutional network to predict the probability of Barrett's esophagus and the trend of microbiota spread. Based on the predicted probability of Barrett's esophagus and the trend of microbiota spread, the lesion risk level is calculated using a microbiota loss quantification algorithm. Combining the patient's gastric acid reflux frequency and esophageal pH fluctuation data, a microbiota-environment interaction model is used to simulate the microbiota colonization coefficient, and the sensitivity parameters are corrected in real time. The corrected sensitivity parameters are transmitted to the clinical diagnostic platform to perform multi-dimensional cancer risk extrapolation, trigger graded diagnosis recommendations, and generate a visual report of the risk heatmap.

[0080] Among them, the microbial community dynamic evolution model is an analytical model that describes or predicts the dynamic changes in the composition, structure and function of the microbial community over time and under the influence of factors such as environment and host status, and integrates multi-dimensional data.

[0081] It is understood that the microbial dynamic evolution model in this application, by integrating microbial-host association maps and spatiotemporal data, accurately predicts the probability of Barrett's esophagus and the trend of microbial spread, providing a dynamic basis for lesion risk assessment. Its ability to integrate multi-dimensional data supports the effective operation of microbial loss-of-measure algorithms and microbial-environment interaction models, simulating microbial colonization coefficients and real-time correction of sensitivity parameters, improving the accuracy and timeliness of disease risk prediction, providing dynamically updated key parameters for clinical diagnostic platforms, and promoting multi-dimensional carcinogenesis risk extrapolation that is more aligned with patients' actual situations. Through tiered diagnostic recommendations and visual reports, it enhances the scientific rigor and relevance of clinical decision-making.

[0082] It should be noted that the microbial community dynamic evolution model:

[0083]

[0084]

[0085] in, Let t be the state vector of the bacterial community; This is the adjacency matrix of the microbial community-host association map; For host state parameters; This represents a vector of environmental factors. For ST-GCN network parameters; For random disturbance terms; This is the state vector of the bacterial community at time t+1; Let t+1 be the probability of Barrett's esophagus occurring. It is the sigmoid activation function; , These are the parameters for the probability prediction layer.

[0086] Formula for spatiotemporal graph convolutional networks:

[0087]

[0088] in, Let be the feature matrix of the graph nodes at time t+1; Let k be the adjacency matrix of the graph. These are the convolution weights corresponding to the d-th feature dimension; Temporal convolution kernel; This represents the number of spatial convolution kernels; For feature dimensions; For activation functions; This is a temporal convolution operation; Let be the d-th dimension characteristic matrix of the nodes in the graph at time t.

[0089] Formula for the microbiome-environment interaction model:

[0090]

[0091] in, Let be the colonization coefficient of the microbial community at time t; This represents the state vector of the bacterial community. This represents a vector of environmental factors. For host state parameters; This is the weight matrix of the bacterial community; This is the weight matrix for the environment; This is the weight matrix for microbial community-environment interactions; This is the weight matrix for the host factors; b is the activation function; b is the bias term. This is a random disturbance term.

[0092] Microbial community loss quantification algorithm formula:

[0093]

[0094] in, The risk level of microbial community loss at time t is used for quantitative analysis. Let be the feature vector of the bacterial community at time t; This represents the baseline vector of gut microbiota characteristics in healthy individuals. To measure the difference between the two; This represents the current microbial diversity index; The average diversity index of healthy individuals; The degree of diversity imbalance; This is the predicted probability of Barrett's esophagus. This is an index indicating the trend of bacterial community spread. , , , which are weighting coefficients, respectively measuring the contribution of deviations in microbial community composition, abnormal diversity, and disease-associated risk.

[0095] For example, such as Figure 4As shown, in the Barrett's esophagus risk assessment, the microbiota dynamic evolution model, based on the microbiota-host association map, integrates esophageal microbiota abundance, spatial distribution, and host epithelial cell lesion characteristics through a spatiotemporal graph convolutional network to predict the probability of Barrett's esophagus occurring in patients within 6 months (e.g., 32%) and the spread path of specific pathogens (e.g., Helicobacter pylori). Combining this prediction result, the microbiota loss-weighting algorithm calculates the moderate lesion risk level. Then, by linking the patient's gastric acid reflux frequency (5 times per week) and esophageal pH fluctuation data, the microbiota-environment interaction model simulates the microbiota colonization coefficient (0.68), and the sensitivity parameters are adjusted in real time (e.g., adjusting the pH influence weight from 0.2 to 0.35). Finally, dynamic parameters are provided to the clinical diagnostic platform to deduce the "needs endoscopic re-examination" classification recommendation and generate a heat map report of the superposition of microbiota dense areas and lesion risk in the upper esophagus.

[0096] In this embodiment, generating a visualization report of a risk heatmap includes: constructing a microbiota-carcinogenesis co-diffusion model; based on the microbiota-carcinogenesis co-diffusion model, combining an improved particle swarm optimization algorithm to simulate the invasion trajectory of pathogenic bacteria in the esophageal mucosa, and mapping the spatial association between the microbiota aggregation area and the epithelial cell dysplasia area to a carcinogenesis probability value through a Gaussian mixture model, generating a carcinogenesis risk probability cloud map; superimposing the carcinogenesis risk probability cloud map onto a three-dimensional esophageal anatomical model, and characterizing the intensity of carcinogenesis risk through color gradients, generating a heatmap visualization report that labels key path nodes of microbiota pathogenicity, the expected carcinogenesis progression time, and the coordinates of high-risk areas.

[0097] Among them, the improved particle swarm optimization algorithm is an intelligent optimization algorithm based on the standard particle swarm optimization algorithm. It optimizes the particle update strategy, enhances the local search capability, or introduces an adaptive parameter adjustment mechanism to more accurately simulate the complex optimization process such as the invasion trajectory of pathogenic bacteria in the esophageal mucosa.

[0098] Understandably, this application's embodiments utilize an improved particle swarm optimization algorithm to accurately simulate the invasion trajectory of pathogenic bacteria in the esophageal mucosa, capturing subtle path changes during bacterial diffusion and avoiding the pitfalls of standard algorithms that easily get trapped in local optima. Simultaneously, the adaptive parameter adjustment mechanism dynamically adapts to the heterogeneity of the esophageal mucosal microenvironment, improving the realism and accuracy of the trajectory simulation. This provides reliable trajectory data for subsequent Gaussian mixture model mapping of the spatial correlation between bacterial aggregation and epithelial cell dysplasia, making the generated cancer risk probability cloud map more closely resemble physiological reality. This ensures that the annotation of key pathogenic path nodes, predicted cancer progression time, and high-risk area coordinates in the heatmap superimposed on the esophageal three-dimensional anatomical model is more accurate, enhancing the reference value of the risk heatmap visualization report for clinical diagnosis and enabling doctors to efficiently identify high-risk cancer areas.

[0099] It should be noted that the formula for the improved particle swarm optimization algorithm is as follows:

[0100]

[0101]

[0102]

[0103] in, Let be the velocity of the i-th particle at time t; Let i be the position of the i-th particle at time t; Let be the inertia weight at the t-th iteration; Let be the velocity of the i-th particle at time t+1; , For learning factors; , It is a random number; This is the best position found so far for the i-th particle; The globally optimal position; For local search correction terms; This is a correction factor; Let be the position of the i-th particle at time t+1; This represents the upper limit of the inertia weight. This is the lower limit of the inertia weight; This represents the maximum number of iterations. For time.

[0104] Formula for the gut microbiota-carcinogenesis co-diffusion model:

[0105]

[0106] in, denoted as the rate of change of pathogen concentration at spatial location x and time t over time. The diffusion coefficient of the bacterial community; For the Laplace operator; This represents the baseline growth rate of the bacterial community. The environmental carrying capacity of the microbial community; The coefficient of synergistic effect between gut microbiota and carcinogenesis; The concentration of pathogenic bacteria at spatial location x and time t; The density of epithelial atypical proliferative cells at spatial location x and time t; This represents the natural decay rate of the bacterial community. The time-varying rate of change of epithelial dysplastic cell density at spatial location x and time t; Let x be the number of tumor cells at spatial location x and time t; The diffusion coefficient of cancerous cells; This represents the baseline growth rate of cancerous cells. The coefficient of synergistic effect between gut microbiota and carcinogenesis; The natural decay rate of cancerous cells; The self-interaction coefficient of cancer cells; This is a density-dependent term for cancerous cells.

[0107] Gaussian mixture model formula:

[0108]

[0109] in, This represents the cancer probability density corresponding to a spatial location x in the esophageal mucosa. The number of Gaussian components in the mixture; It is the k-th Gaussian distribution component; The weight of the k-th Gaussian component; Let be the probability density function of the k-th Gaussian distribution; Let be the mean vector of the k-th Gaussian distribution; Let be the covariance matrix of the k-th Gaussian distribution.

[0110] For example, in intelligent detection of intestinal flora fluorescence microscopy images, the improved particle swarm optimization (PSO) algorithm achieves precise localization and feature extraction of dense flora regions by integrating adaptive inertial weights and a neighborhood particle information correction mechanism. Addressing issues such as blurred edges between flora and epithelial cells, uneven fluorescence signals, and clumping interference from local aggregations in the images, the algorithm dynamically adjusts its particle search strategy: in the global phase, it rapidly traverses the image with a large inertial weight (linearly decaying from 0.9 to 0.4) to capture candidate regions where flora may be distributed; in the local phase, it introduces a neighborhood particle cooperation mechanism, calculating the Euclidean distance between particles to correct the velocity parameter, thus enhancing the detailed recognition of small flora contours (such as cocci with a diameter <5μm). Experimental data show that compared to the standard PSO algorithm, its Dice coefficient for flora region segmentation is improved by 8.3%, the detection rate of low-fluorescence-intensity flora is increased by 12.7%, and the processing time for a single 4K resolution image is reduced to 0.8 seconds. The algorithm outputs precise spatial coordinates and morphological parameters of the bacterial community, providing reliable data support for subsequent calculation of bacterial community density gradients and analysis of spatial correlation with host cells, significantly improving the detection efficiency of abnormal bacterial community aggregation patterns in clinical samples.

[0111] In this embodiment, a microbiota-environment interaction model is used to simulate the microbiota colonization coefficient and correct sensitivity parameters in real time. This includes: constructing a microbiota-environment interaction dynamic model; inputting data on gastric acid reflux frequency, esophageal pH fluctuations, and microbiota metabolite concentrations into the microbiota-environment interaction dynamic model, and simulating the change curve of the microbiota colonization coefficient with environmental factors through differential equations; and correcting sensitivity parameters in real time based on the change curve to control the error between predicted values ​​and clinically measured values.

[0112] Among them, the colonization coefficient is an indicator that quantifies the ability of a microbial community to adapt, reproduce and maintain a stable presence in a specific host environment (such as esophageal mucosa, intestinal tissue, etc.), reflecting the intensity of its interaction with environmental factors and host status.

[0113] It is understood that the embodiments of this application integrate data such as the frequency of gastric acid reflux, pH fluctuations, and the concentration of metabolites through a microbiota-environment interaction dynamic model, and use differential equations to dynamically present the pattern of its changes with the environment, providing a quantitative basis for analyzing the interaction mechanism between the microbiota and the host microenvironment. The sensitivity parameters are corrected in real time through Bayesian optimization algorithm, reducing the error between the predicted value and the clinical measured value, improving the model's prediction accuracy of the microbiota colonization status, and enhancing the scientificity and accuracy of medical decision-making.

[0114] It should be noted that the formula for the differential equation is:

[0115]

[0116] in, The rate of change of the colonization coefficient of the microbial community over time; This represents the intrinsic growth rate of the bacterial community in an ideal environment. The maximum carrying capacity of the environment for microbial colonization; This represents the colonization coefficient of the microbial community. For logistic growth terms; The inhibition coefficient of gastric acid reflux on bacterial colonization; This refers to the frequency of gastric acid reflux. This item represents the inhibitory effect on gastric acid reflux. The pH deviation represents the inhibition coefficient of bacterial colonization. This refers to the real-time pH value within the esophagus. The optimal pH value for bacterial growth; This is the term representing the inhibitory effect of pH deviation; The coefficient representing the promotion effect of low concentrations of metabolites on bacterial colonization. The concentration of bacterial community metabolites; This refers to the promoting effect of low concentrations of metabolites; The inhibition coefficient of high concentrations of metabolites on bacterial colonization; This refers to the inhibitory effect of high concentrations of metabolites.

[0117] Based on the curve of the colonization coefficient output by the microbiome-environment interaction dynamics model, which shows the changes in environmental factors such as the frequency of gastric acid reflux, esophageal pH fluctuations, and the concentration of microbiome metabolites (i.e., the dynamic trend of the colonization coefficient predicted by the model), the sensitivity parameters in the model that reflect the intensity of the influence of environmental factors on microbiome colonization (such as the inhibition coefficient of gastric acid reflux on colonization, the inhibition coefficient of pH deviation from the optimal value, etc.) can be continuously and dynamically adjusted using algorithms such as Bayesian optimization. Through this real-time parameter correction, the deviation between the microbiome colonization coefficient predicted by the model (the predicted value obtained from the change curve) and the microbiome colonization coefficient actually detected in clinical practice (the measured value) can be continuously reduced, making the model's simulation of the colonization state of microbiome in the host environment closer to the real physiological scenario, thereby providing more accurate quantitative support for microbiome-related disease risk assessment and clinical diagnosis and treatment decisions.

[0118] For example, in the intelligent detection of microbial images in reflux esophagitis, a microbial-environment interaction dynamic model is first used to integrate data on the frequency of gastric acid reflux (e.g., 3 times daily), the fluctuation range of esophageal pH (2.1-6.8), and the concentration of microbial metabolites. Differential equations are then used to simulate the change curve of the colonization coefficient of the lower esophagus (0.32-0.78) with acid exposure time. Based on this curve, the image detection algorithm parameters are optimized to accurately identify microbial clusters (e.g., streptococcal clumps with a diameter >20 μm) that match the high colonization coefficient region in fluorescence microscopy images, and associate them with mucosal congestion and edema areas. This enables intelligent quantitative detection of abnormal microbial colonization in patients with reflux esophagitis, providing a microbiological basis for assessing the degree of inflammation.

[0119] In this embodiment, a spatiotemporal graph convolutional network is used to construct a microbial community dynamic evolution model to obtain the probability of Barrett's esophagus and the microbial community diffusion trend. This includes: constructing a microbial community dynamic evolution model; based on the microbial community dynamic evolution model, setting the high-incidence areas of precancerous lesions in the dentate line of the lower esophagus and the cardia mucosal folds as core nodes of the graph network, and combining the microscopic characteristics of microbial biofilm thickness, epithelial cell apoptosis rate, and inflammatory factor concentration, modeling the spatiotemporal correlation between microbial invasion depth and the degree of epithelial dysplasia through a spatiotemporal graph convolutional network to predict the probability of Barrett's esophagus and the corresponding microbial community diffusion trend.

[0120] Among them, the microbial community dynamic evolution model describes the dynamic process of changes in the structure and quantity of microbial communities in the host environment with time, space and environmental factors, reveals the interaction law between microbial communities and the host and environment and predicts their evolutionary trend.

[0121] It is understood that the embodiments of this application utilize a microbial community dynamic evolution model, integrating spatial features of key anatomical regions in the lower esophagus (such as the dentate line and cardia mucosal folds) with microscopic dynamic information such as microbial biofilm thickness and epithelial cell apoptosis rate through a spatiotemporal graph convolutional network. This model accurately models the spatiotemporal correlation between microbial community invasion and epithelial dysplasia, quantitatively predicts the probability of Barrett's esophagus, clearly presents the trend of microbial community diffusion, provides a visualized dynamic basis for analyzing microbial community-driven precancerous lesions, and provides scientific support for early clinical screening of high-risk groups and the development of targeted intervention strategies, thereby improving the accuracy and foresight of Barrett's esophagus prevention and treatment.

[0122] In this embodiment, the triggering of graded diagnostic recommendations includes: setting up a four-level diagnostic mechanism. When the probability of Barrett's esophagus is less than 20%, with no abnormal spread of the flora and normal epithelial cells, it is classified as a level 1 low-risk condition, and dietary intervention and follow-up every 6 months are recommended. When the probability of occurrence is 20%-50%, with localized flora aggregation and mild epithelial hyperplasia, it is classified as a level 2 low-to-medium risk condition, and probiotic intervention and gastroscopy follow-up are recommended. When the probability of occurrence is 50%-80%, with flora spreading to the upper esophagus and accompanied by moderate dysplasia, it is classified as a level 3 medium-to-high risk condition, triggering endoscopic mucosal resection evaluation and targeted antimicrobial therapy. When the probability of occurrence is greater than 80%, with flora invasion of the entire esophagus and accompanied by severe dysplasia or early cancerous changes, it is classified as a level 4 high risk condition, initiating multidisciplinary consultation and radical surgery evaluation.

[0123] It is understood that the embodiments of this application use a four-level mechanism to precisely stratify the risk of Barrett's esophagus, matching differentiated intervention plans based on the probability of occurrence, the status of bacterial flora spread, and the degree of epithelial lesions: low risk focuses on dietary control and regular monitoring, medium and low risk strengthens probiotic intervention and gastroscopy follow-up, medium and high risk triggers endoscopic resection assessment and targeted antimicrobial therapy, and high risk initiates multidisciplinary consultation and radical surgery assessment. This avoids overtreatment of low-risk individuals and delayed intervention for high-risk individuals, and improves the targeting and effectiveness of intervention through tiered measures. At the same time, it standardizes the follow-up frequency and treatment pathway, helps clinicians to efficiently allocate medical resources, and reduces the risk of Barrett's esophagus progressing to esophageal cancer.

[0124] According to the embodiments of this application, an intelligent detection method for esophageal microbiota images for predicting the risk of gastroesophageal reflux disease is proposed. This method integrates esophageal mucosal microscopic images with multi-dimensional clinical data, accurately quantifies the spatial displacement characteristics of the microbiota using an image registration algorithm, and achieves efficient extraction of microscopic features of the microbiota by combining an improved U-Net++ network. Furthermore, it deeply integrates the dynamic changes of the microbiota with epithelial cell lesion features through a spatiotemporal attention mechanism. The generated microbiota-host association map breaks the separation between microbiota data and mucosal damage information in traditional detection, and realizes cross-scale association analysis from microscopic microbiota behavior to macroscopic lesions, significantly improving the ability to capture early microbial abnormalities and occult mucosal damage. A microbial dynamic evolution model constructed based on a spatiotemporal graph convolutional network can accurately predict the probability of Barrett's esophagus and the trend of microbial diffusion. Combined with environmental data such as the frequency of gastric acid reflux and esophageal pH fluctuations, the model uses a microbial-environment interaction model to correct detection sensitivity parameters in real time. This effectively solves the problem of poor adaptability to individual differences and environmental fluctuations in traditional methods, making risk assessment more closely aligned with the patient's actual pathological state and significantly improving the accuracy and timeliness of precancerous lesion early warning. The final generated tiered diagnostic recommendations and risk heatmap visualization report provide clinicians with intuitive decision-making basis from the pathogenesis pathway of microbial flora to the risk of cancer. This not only shortens the detection cycle (by more than 60% compared to traditional culture methods) but also reduces the risk of overtreatment and missed diagnosis through precise stratified treatment recommendations. It comprehensively improves the accuracy, clinical suitability, and diagnostic efficiency of reflux esophagitis and related precancerous lesion detection, providing new technical support for the early intervention and precise management of digestive tract diseases. Thus, it solves the problems of long detection cycles and weak microscopic analysis capabilities in existing technologies.

[0125] The following specific embodiment will illustrate an intelligent detection method for esophageal microbiota images for predicting the risk of gastroesophageal reflux disease (GERD). Figure 5 As shown, it includes:

[0126] One hundred and twenty patients with reflux esophagitis (72 males and 48 females, aged 23-72 years, mean age 46.3±9.8 years) and sixty healthy controls (32 males and 28 females, aged 22-69 years, mean age 43.5±10.1 years) were selected. All participants signed informed consent forms and the study was ethically approved. Samples were collected using 2.8mm sterile disposable biopsy forceps in the lower esophagus, 2-5 cm from the dentate line. Six to eight typical lesion points (congestion and erosion) were selected in the patient group, while three to four normal mucosal points were taken from the control group. Esophageal tissue samples were immediately placed in sterile cryovials containing RNA preservation solution, labeled, and stored at -80°C. Temperature was continuously monitored during transport (fluctuations ≤±2°C). For sample processing, after fixation with 4% paraformaldehyde for 24 h, the samples were dehydrated in a gradient of ethanol (70%, 80%, 95%, 100%) for 15 min each, cleared twice with xylene (10 min each time), and embedded in paraffin to prepare 5 μm serial sections. Multiplex FISH staining was performed: after dewaxing, the sections were incubated with 3% H2O2 at room temperature for 10 min to inactivate endogenous peroxidase, followed by antigen retrieval at high temperature with 0.1 mol / L citrate buffer (pH 6.0) for 15 min, and then a mixture containing probes such as EUB338-Cy3 (total bacteria), Lacto-FITC (Lactobacillus), and Strepto-Cy5 (Streptococcus) (10 ng / μL concentration of each probe) was added. Hybridization was performed at 37°C for 16 h in a humidified chamber, followed by washing twice with 2×SSC solution at 37°C (10 min each time), nucleus staining with DAPI for 5 min, and mounting with an anti-fluorescence quencher. Images were acquired using a Zeiss LSM900 confocal microscope. DAPI (nuclei appear blue) was excited by a 405nm laser, FITC (lactobacteria appear green) by a 488nm laser, Cy3 (total bacteria appear red) by a 555nm laser, and Cy5 (streptococci appear magenta) by a 640nm laser. Z-stack scanning was performed at 40x oil immersion with a 0.3μm step. 15-20 non-overlapping fields were acquired per sample, and images were saved as 1024×1024 pixel TIFF format, with acquisition parameters (laser intensity, gain, etc.) included. Clinical data were extracted from the electronic medical record system: basic information (age, gender, BMI, etc.), symptom data (frequency of heartburn, VAS score for acid reflux), 24-hour esophageal pH monitoring results (Medtronic Digitrapper MKIV device, recording reflux frequency, percentage of total time with pH < 4, and longest reflux duration), Los Angeles classification (AD level), and past medication history (PPI usage duration and dosage, etc.).

[0127] Systematic preprocessing of microscopic images of bacterial colonies was performed: First, nonlocal mean filtering was used for noise reduction, with a search window of 7×7, a similarity window of 3×3, and a filtering coefficient of 0.1, reducing detail loss by 30% compared to traditional Gaussian filtering; then, the CLAHE algorithm was used to enhance contrast, with a size of 8×8 and a limiting threshold of 0.02, increasing the grayscale difference between the bacterial colonies and the background to 1.5 times that of the original image; finally, pixel values ​​were linearly normalized to the [0,1] interval to eliminate brightness differences between devices. Clinical data preprocessing included: continuous variables (such as reflux frequency) were mapped to [0,1] using Min-Max normalization, categorical variables (such as Los Angeles classification) were converted into 4-dimensional vectors using one-hot encoding, and missing values ​​(<5%) were filled using the KNN algorithm (k=5) to ensure data integrity. The perceptual network is constructed based on the PyTorch 1.13 framework: The input layer adopts a dual-branch structure, receiving a 512×512 resolution image (after downsampling) and a 12-dimensional clinical feature vector respectively; The feature extraction layer contains 4 residual blocks, each consisting of two 3×3 convolutional layers (stride 2, padding=1), a BatchNorm layer, and a ReLU activation function, progressively compressing the image features to 32×32×512 dimensions; The fusion layer merges the image features and clinical features (mapped to 512 dimensions via a fully connected layer) into a 32×32×1024 tensor through feature concatenation. Image registration employs the SIFT+RANSAC algorithm: For sequential images of different biopsy sites from the same patient, feature points are first extracted using the SIFT algorithm (detection threshold 0.04, edge threshold 10), yielding an average of 300-500 stable feature points per image; then, a coarse matching is performed using a FLANN matcher (kd-tree index, matching distance ratio 0.7), followed by RANSAC algorithm (1000 iterations, threshold 1.5 pixels) to remove mismatched points (retention rate approximately 60%). The affine transformation matrix is ​​calculated, outputting the translation (μm) and rotation angle (°) of the bacterial community along the x / y axes, quantifying spatial displacement features. The improved U-Net++ network optimizes the original: a coordinate attention module is added to the skip connections (using 1×1 convolution to compress channels and generate x / y axis attention weights) to enhance attention to densely populated bacterial regions; the deep network uses dilated convolutions (rate 2) to expand the receptive field to 64×64 pixels without reducing resolution. During model training, the dataset was divided into a training set (108 patients + 54 controls), a validation set (13 patients + 6 controls), and a test set (13 patients + 6 controls) in an 8:1:1 ratio. The optimization objective was Dice loss (weight 0.3) + cross-entropy loss (weight 0.7). The Adam optimizer had an initial learning rate of 0.0001, which decayed by 50% every 20 epochs. After 150 epochs of training, the Dice coefficient on the validation set reached 0.92 ± 0.03.Three types of features were extracted from the segmentation results: bacterial density (number of bacteria per mm², calculated by pixel count), aggregation degree (Moran's I index, ranging from -1 to 1, >0 indicates aggregated distribution), and fluorescence intensity gradient (Sobel operator calculates the gradient in the x / y directions, and the mean value represents the fluorescence change rate). A spatiotemporal attention mechanism was used to fuse the features: in the temporal dimension, a GRU network (128 hidden layers) was used to model the sequence features, generating temporal attention weights (highlighting time points with dramatic changes in bacterial dynamics); in the spatial dimension, features were compressed to a single channel using 1×1 convolution, and Sigmoid activation was used to generate a spatial attention map (enhancing epithelial lesion areas); after weighted fusion, a 256×256×3 bacterial-host association map was output (red channel: bacterial density, green channel: host cell lesion severity, blue channel: interaction strength).

[0128] A spatiotemporal graph convolutional network (ST-GCN) is constructed based on the microbiome-host association map: Nodes are defined as microbiome clusters (connected regions containing >5 bacteria) and epithelial cell clusters (cell aggregation regions with an area >100μm²). Each node's feature vector contains 16 parameters (microbiome features: density, aggregation degree, etc., 8 items; host features: cell morphology features, nucleocytoplasmic ratio, etc., 8 items). Edge weights are jointly determined by Euclidean distance (<50μm, set to 1; otherwise, 0) and interaction strength (based on the spatial adjacency strength between microbiome and host cells, >0.6, set to 1), constructing an adjacency matrix. The network contains 3 spatiotemporal blocks: each block consists of a graph convolutional layer (using the ChebNet operator, K=2nd order polynomial) and a 1D convolutional layer (kernel size 3, stride 1), capturing spatial dependence and temporal dynamics respectively; the output layer predicts the probability of Barrett's esophagus (0-1) and the microbiome diffusion vector (velocity μm / h, orientation angle °) through a fully connected layer. The model was trained using the Adam optimizer (learning rate 0.001) with a loss function of MSE (probability prediction) + cross-entropy (diffusion direction classification). After 80 training epochs, the prediction accuracy on the test set was 89.6%, and the AUC was 0.91. The steps of the microbial imbalance quantification algorithm were: ① Shannon diversity index (H = -Σpi × lnpi, where pi is the proportion of the i-th microbial group); ② Dominant microbial proportion (relative abundance of the top-ranked microbial group, >30% defined as imbalance); ③ Host interaction deviation (Euclidean distance between the current sample and the healthy group's interaction strength); these three factors were weighted at a ratio of 0.4:0.3:0.3 to obtain the imbalance index, which was used to classify risk levels (<0.25 low risk, 0.25-0.65 medium risk, >0.65 high risk), combined with Barrett's esophageal prediction probability (the level was increased by 1 level when >50%). The microbiota-environment interaction model was fitted using multiple linear regression: colonization coefficient = 0.32 × reflux frequency (times / 24h) + 0.28 × pH fluctuation range (pH range within 24h) + 0.15 × Los Angeles grade score (A=1, B=2, C=3, D=4) - 0.42 (R²=0.78, P<0.001). For every 0.1 increase in the coefficient, the model sensitivity (true positive rate) increased by 5%. The corrected sensitivity parameters were transmitted to a SpringBoot-based clinical diagnostic platform via a RESTful API. The platform includes a data layer (MySQL database storing images and clinical data), a model layer (deploying algorithms such as ST-GCN), and an application layer (visual interface). The multidimensional cancer risk projection uses a random forest model (100 decision trees). The input features include 15 items such as imbalance index, Barrett probability, and colonization coefficient. The output is a risk value (0-100) and triggering classification suggestions: low risk (<30 points): 6-month follow-up and dietary intervention; medium risk (30-70 points): standard dose PPI treatment and 3-month follow-up; high risk (>70 points): endoscopic radiofrequency ablation and 1-month follow-up.Risk heatmaps were generated using Matplotlib, dividing the lower esophagus into six regions (2-3 cm, 3-4 cm, and 4-5 cm from the dentate line, anterior / posterior wall). RGB pseudo-color (red = high risk, yellow = medium risk, green = low risk) was used to display the risk value of each region. The report included a supplementary microbial distribution heatmap, dynamic evolution curves (comparison of the last three tests), and personalized intervention suggestions (such as adjusting PPI dosage and probiotic supplementation). Validated with 200 independent samples (150 patients + 50 controls), the risk grading and pathological diagnosis concordance rate was 87.5%, with a heatmap AUROC of 0.89, significantly higher than traditional pathological diagnosis (76.3%, P<0.01). The average diagnosis time was reduced from 48 hours to 2 hours.

[0129] In summary, this invention ensures data quality through standardized sample collection and processing procedures, accurately extracts microbial characteristics by combining advanced algorithms such as nonlocal mean filtering and improved U-Net++, efficiently predicts the probability of Barrett's esophagus and the trend of microbial spread using the ST-GCN model, optimizes risk assessment through microbial dysbiosis quantification and environment interaction models, and finally generates accurate grading suggestions and visualized heatmaps through a clinical diagnostic platform. This not only improves the concordance rate between risk grading and pathological diagnosis, outperforming traditional pathological diagnosis, but also shortens the average diagnosis time and provides personalized intervention plans, offering strong support for efficient and accurate diagnosis and clinical decision-making regarding reflux esophagitis and cancer risk.

[0130] Next, referring to the accompanying drawings, an intelligent esophageal microbiota image detection system for predicting the risk of gastroesophageal reflux disease is described according to an embodiment of this application.

[0131] Figure 6 This is a schematic diagram of the structure of an intelligent esophageal microbiota image detection system for predicting the risk of gastroesophageal reflux disease, according to an embodiment of this application.

[0132] like Figure 6 As shown, the intelligent esophageal microbiota image detection system 10 for predicting the risk of gastroesophageal reflux disease includes: an acquisition module 100, an extraction module 200, and a generation module 300.

[0133] The acquisition module 100 is used to acquire microscopic images of the microbial community and clinical feature data of esophageal mucosal samples; the extraction module 200 is used to construct a perceptual network based on the microscopic images of the microbial community and clinical feature data, and based on the perceptual network, use an image registration algorithm to quantify the spatial displacement features of the microbial community in the esophageal sample sequence, and simultaneously use an improved U-Net++ network to segment the image, extract microbial community density, aggregation degree and fluorescence intensity gradient features, and fuse microbial community dynamics and epithelial cell lesion features through a spatiotemporal attention mechanism to generate a microbial community-host association map; the generation module 300 is used to generate a microbial community-host association map based on the microbial community-host association. The microbial community dynamic evolution model is constructed using a spatiotemporal graph convolutional network to predict the probability of Barrett's esophagus and the trend of microbial community spread. Based on the predicted probability of Barrett's esophagus and the trend of microbial community spread, the risk level of the lesion is calculated through a microbial community loss-of-measure algorithm. Combined with the patient's gastric acid reflux frequency and esophageal pH fluctuation data, a microbial community-environment interaction model is used to simulate the microbial community colonization coefficient, and the sensitivity parameters are corrected in real time. The corrected sensitivity parameters are transmitted to the clinical diagnostic platform to perform multi-dimensional cancer risk extrapolation, trigger graded diagnosis suggestions, and generate a visual report of risk heatmap.

[0134] It should be noted that the foregoing explanation of an embodiment of an intelligent detection method for esophageal microbiota images for predicting the risk of gastroesophageal reflux disease also applies to the intelligent detection system for esophageal microbiota images for predicting the risk of gastroesophageal reflux disease in this embodiment, and will not be repeated here.

[0135] According to the embodiments of this application, an intelligent esophageal microbiota image detection system for predicting the risk of gastroesophageal reflux disease is proposed. By integrating esophageal mucosal microscopic images of microbiota with multi-dimensional clinical data, the system accurately quantifies the spatial displacement characteristics of microbiota using image registration algorithms, and achieves efficient extraction of microscopic features of microbiota by combining an improved U-Net++ network. Furthermore, by deeply fusing microbiota dynamic changes and epithelial cell lesion characteristics through a spatiotemporal attention mechanism, the generated microbiota-host association map breaks the separation between microbiota data and mucosal damage information in traditional detection. It realizes cross-scale association analysis from microscopic microbiota behavior to macroscopic lesions, significantly improving the ability to capture early microbial abnormalities and occult mucosal damage. A microbial dynamic evolution model constructed based on a spatiotemporal graph convolutional network can accurately predict the probability of Barrett's esophagus and the trend of microbial diffusion. Combined with environmental data such as the frequency of gastric acid reflux and esophageal pH fluctuations, the model uses a microbial-environment interaction model to correct detection sensitivity parameters in real time. This effectively solves the problem of poor adaptability to individual differences and environmental fluctuations in traditional methods, making risk assessment more closely aligned with the patient's actual pathological state and significantly improving the accuracy and timeliness of precancerous lesion early warning. The final generated tiered diagnostic recommendations and risk heatmap visualization report provide clinicians with intuitive decision-making basis from the pathogenesis pathway of microbial flora to the risk of cancer. This not only shortens the detection cycle (by more than 60% compared to traditional culture methods) but also reduces the risk of overtreatment and missed diagnosis through precise stratified treatment recommendations. It comprehensively improves the accuracy, clinical suitability, and diagnostic efficiency of reflux esophagitis and related precancerous lesion detection, providing new technical support for the early intervention and precise management of digestive tract diseases. Thus, it solves the problems of long detection cycles and weak microscopic analysis capabilities in existing technologies.

[0136] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include:

[0137] The memory 701, the processor 702, and the computer program stored on the memory 701 and executable on the processor 702.

[0138] When the processor 702 executes the program, it implements the intelligent detection method for esophageal flora images for predicting the risk of gastroesophageal reflux disease provided in the above embodiments.

[0139] Furthermore, electronic devices also include:

[0140] Communication interface 703 is used for communication between memory 701 and processor 702.

[0141] The memory 701 is used to store computer programs that can run on the processor 702.

[0142] The memory 701 may include high-speed RAM (Random Access Memory) and may also include non-volatile memory, such as at least one disk storage device.

[0143] If the memory 701, processor 702, and communication interface 703 are implemented independently, then the communication interface 703, memory 701, and processor 702 can be interconnected via a bus to complete communication between them. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 7 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0144] Optionally, in a specific implementation, if the memory 701, processor 702, and communication interface 703 are integrated on a single chip, then the memory 701, processor 702, and communication interface 703 can communicate with each other through an internal interface.

[0145] The processor 702 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of this application.

[0146] In the description of this specification, the references to "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0147] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0148] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0149] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any of the following techniques known in the art, or a combination thereof: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0150] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method for intelligent detection of esophageal microbiota images for predicting the risk of gastroesophageal reflux disease, characterized in that, include: Obtain microscopic images of the microbial community and clinical characteristics of esophageal mucosal samples; Based on the microscopic images of the microbial community and clinical feature data, a perception network is constructed. Based on the perception network, an image registration algorithm is used to quantify the spatial displacement features of the microbial community in the esophageal sample sequence. At the same time, an improved U-Net++ network is used to segment the image and extract the microbial community density, aggregation degree and fluorescence intensity gradient features. The microbial community dynamics and epithelial cell lesion features are fused through a spatiotemporal attention mechanism to generate a microbial community-host association map. Based on the aforementioned microbiota-host association map, a dynamic evolution model of the microbiota is constructed using a spatiotemporal graph convolutional network to predict the probability of Barrett's esophagus and the trend of microbiota spread. This includes: constructing a dynamic evolution model of the microbiota; based on the aforementioned dynamic evolution model of the microbiota, setting the high-incidence areas of precancerous lesions in the dentate line of the lower esophagus and the folds of the cardia mucosa as core nodes of the graph network; combining the microscopic characteristics of microbiota biofilm thickness, epithelial cell apoptosis rate, and inflammatory factor concentration; and modeling the spatiotemporal correlation between microbiota invasion depth and the degree of epithelial dysplasia using a spatiotemporal graph convolutional network to predict the probability of Barrett's esophagus and the corresponding trend of microbiota spread. Based on the predicted probability of Barrett's esophagus and the trend of bacterial spread, the risk level of the lesion is calculated by a bacterial dystrophic quantification algorithm. Combined with the patient's gastric acid reflux frequency and esophageal pH fluctuation data, a bacterial colonization coefficient is simulated using a bacterial-environment interaction model. The sensitivity parameters are corrected in real time, and the corrected sensitivity parameters are transmitted to the clinical diagnostic platform for multi-dimensional cancer risk extrapolation, triggering graded diagnostic recommendations and generating a visual report of the risk heatmap.

2. The intelligent detection method for esophageal microbiota images for predicting the risk of gastroesophageal reflux disease according to claim 1, characterized in that, Generate a visual report of the risk heatmap, including: Constructing a microbiome-carcinogenesis co-diffusion model; Based on the aforementioned microbiome-carcinogenesis co-diffusion model, combined with an improved particle swarm optimization algorithm to simulate the invasion trajectory of pathogenic bacteria in the esophageal mucosa, the spatial correlation between the microbiome aggregation region and the epithelial cell atypical proliferation region is mapped to a carcinogenesis probability value through a Gaussian mixture model, generating a carcinogenesis risk probability cloud map. The cancer risk probability cloud map is overlaid onto the esophageal three-dimensional anatomical model. The intensity of cancer risk is represented by color gradients, and a heat map visualization report is generated, which marks the key path nodes of bacterial flora, the expected time of cancer progression, and the coordinates of high-risk areas.

3. The intelligent detection method for esophageal microbiota images for predicting the risk of gastroesophageal reflux disease as described in claim 1, characterized in that, The spatiotemporal attention mechanism uses the following formula: ； in, Output results for attention; This is the microbial community dynamics feature matrix; A matrix representing the characteristics of epithelial cell lesions; The query vector generated by M; The key vector generated by E; Key vector The transpose of the matrix; The dimension of the key vector; It is a value vector; The scaled score represents the microbiome-host feature association score.

4. The intelligent detection method for esophageal microbiota images for predicting the risk of gastroesophageal reflux disease according to claim 1, characterized in that, Using a microbial community-environment interaction model, the microbial community colonization coefficient is simulated, and sensitivity parameters are corrected in real time, including: Construct a dynamic model of microbial community-environment interactions; The frequency of gastric acid reflux, esophageal pH fluctuation data, and concentration data of microbial metabolites are input into the microbial community-environment interaction dynamics model, and the curve of microbial community colonization coefficient changing with environmental factors is simulated through differential equations. Based on the aforementioned change curve, the sensitivity parameter is adjusted in real time to control the error between the predicted value and the clinically measured value.

5. The intelligent detection method for esophageal microbiota images for predicting the risk of gastroesophageal reflux disease according to claim 1, characterized in that, Construct a perceptual network, and based on the perceptual network, use an image registration algorithm to quantify the spatial displacement features of the esophageal sample sequence microbiota, including: Construct a multimodal sensing network; Based on the multimodal perception network, a dynamic weight matrix is ​​generated by fusing the texture features of microscopic images of microbial communities with the structured information of clinical feature data. Based on the dynamic weight matrix, an image registration algorithm is used to calculate the three-dimensional spatial displacement vector of the microbial community in the esophageal sample sequence, and to quantify the migration distance and orientation angle parameters of the microbial community.

6. The intelligent detection method for esophageal microbiota images for predicting the risk of gastroesophageal reflux disease according to claim 1, characterized in that, The improved U-Net++ network was used to segment the image, and the bacterial community density, aggregation degree, and fluorescence intensity gradient features were extracted, including: Construct an improved U-Net++ network; Based on the improved U-Net++ network, pixel-level segmentation of the microscopic image of the bacterial community is performed to generate a mask for the bacterial community region. Based on the community region mask, a density clustering algorithm is used to calculate the community density index, morphological operators are used to extract aggregation characteristics, and the Sobel operator is used to calculate the gradient magnitude and direction characteristics of fluorescence intensity.

7. The intelligent detection method for esophageal microbiota images for predicting the risk of gastroesophageal reflux disease according to claim 1, characterized in that, The improved U-Net++ network formula is as follows: ； in, Let be the image segmentation accuracy at time t; Based on the accuracy of segmentation; The fluorescence contrast of the bacterial community at time t; Standard contrast threshold; This represents the regional weighting coefficient of the microbial community. Noise impact factor; Let be the image noise level at time t.

8. The intelligent detection method for esophageal microbiota images for predicting the risk of gastroesophageal reflux disease according to claim 1, characterized in that, Triggering tiered diagnostic recommendations include: A four-tiered diagnostic mechanism is established. When the incidence of Barrett's esophagus is less than 20%, with no abnormal bacterial spread and normal epithelial cells, it is classified as Level 1 (low risk), with dietary intervention and follow-up every 6 months recommended. When the incidence is 20%-50%, with local bacterial aggregation and mild epithelial hyperplasia, it is classified as Level 2 (low to medium risk), with probiotic intervention and endoscopic follow-up recommended. When the incidence is 50%-80%, with bacterial spread to the upper esophagus and moderate dysplasia, it is classified as Level 3 (high to medium risk), triggering evaluation for endoscopic mucosal resection and targeted antimicrobial therapy. When the incidence is greater than 80%, with bacterial invasion across the entire esophagus and severe dysplasia or early cancerous changes, it is classified as Level 4 (high risk), initiating multidisciplinary consultation and evaluation for radical surgery.

9. An intelligent image detection system for esophageal microbiota for predicting the risk of gastroesophageal reflux disease, characterized in that, include: The acquisition module is used to acquire microscopic images of the microbial community and clinical characteristic data of esophageal mucosal samples; The extraction module is used to construct a perception network based on the microscopic images of the microbial community and clinical feature data. Based on the perception network, an image registration algorithm is used to quantify the spatial displacement features of the microbial community in the esophageal sample sequence. At the same time, an improved U-Net++ network is used to segment the image and extract the microbial community density, aggregation degree and fluorescence intensity gradient features. The microbial community dynamics and epithelial cell lesion features are fused through a spatiotemporal attention mechanism to generate a microbial community-host association map. The generation module is used to construct a dynamic evolution model of the microbiota based on the microbiota-host association map using a spatiotemporal graph convolutional network, predict the probability of Barrett's esophagus and the microbiota spread trend, calculate the lesion risk level based on the predicted probability of Barrett's esophagus and the microbiota spread trend using a microbiota skewing quantification algorithm, combine the patient's gastric acid reflux frequency and esophageal pH fluctuation data, use a microbiota-environment interaction model to simulate the microbiota colonization coefficient, correct the sensitivity parameters in real time, and transmit the corrected sensitivity parameters to the clinical diagnostic platform for multi-dimensional cancer risk extrapolation, triggering graded diagnostic suggestions and generating [the necessary data]. The visualization report of the risk heatmap, which utilizes a spatiotemporal graph convolutional network to construct a microbial community dynamic evolution model to predict the probability of Barrett's esophagus and the trend of microbial community spread, includes: constructing a microbial community dynamic evolution model; based on the microbial community dynamic evolution model, setting the high-incidence areas of precancerous lesions in the dentate line of the lower esophagus and the gastric cardia mucosal folds as core nodes of the graph network, and combining the microscopic characteristics of microbial biofilm thickness, epithelial cell apoptosis rate, and inflammatory factor concentration, modeling the spatiotemporal correlation between microbial community invasion depth and the degree of epithelial dysplasia through a spatiotemporal graph convolutional network, and predicting the probability of Barrett's esophagus and the corresponding microbial community spread trend.

Images