Dialyzer blood coagulation grading intelligent data processing method and system based on multi-source data fusion

Abstract

The invention provides a dialyzer blood coagulation grading intelligent data processing method and system based on multi-source data fusion, and relates to the technical field of data processing, and the method comprises the steps: obtaining dialyzer data; performing preprocessing of region segmentation on the dialyzer data to obtain preprocessed data of a dialyzer fiber bundle structure; analyzing the preprocessed data into a discrete feature data set representing structural features, processing the discrete feature data set to obtain barycentric coordinates of a dialyzer fiber bundle core distribution area, constructing an analysis plane by taking the barycentric coordinates as a reference, and determining a reference analysis framework matched with the dialyzer fiber bundle distribution features on the analysis plane. According to the invention, objectification, automation and clinical suitability of blood coagulation grading of the dialyzer are realized.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to an intelligent data processing method and system for dialyzer coagulation grading based on multi-source data fusion. Background Technology

[0002] Patients with end-stage renal disease rely on hemodialysis to maintain their lives. Post-dialysis coagulation grading assessment is a core step in developing individualized anticoagulation protocols and reducing the risk of treatment interruption. Currently, the clinical practice generally adopts a grading method that relies on medical staff to visually observe the color of fiber bundles and the adhesion of blood clots. However, this grading method has technical limitations in practical application.

[0003] For example, in a patient with diabetic nephropathy and hypertension, after completing a routine dialysis session, scattered dark red blood clots were observed near the arterial end of the dialyzer. Based on their clinical experience, the patient determined this to be a coagulation condition of the corresponding grade and planned to adjust the anticoagulant dosage. However, the nurse responsible for checking the tubing believed that the coagulation range did not meet the corresponding standard and only met the lower grade of coagulation. Due to the lack of objective quantitative evidence in the assessment process, a dispute arose. Although the final judgment was aided by reviewing relevant pressure data from the dialysis process, the timely adjustment of the anticoagulation regimen had been delayed, and the risk of tubing blockage during the patient's next dialysis session had increased. This case highlights the technical shortcomings of the traditional visual method, which is highly subjective, has vague grading standards, lacks objective quantitative support, and cannot be effectively linked with diagnostic and treatment data. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and system for intelligent data processing of dialyzer coagulation grading based on multi-source data fusion, so as to realize the objectivity, automation and clinical adaptability of dialyzer coagulation grading.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: Firstly, a smart data processing method for dialyzer coagulation grading based on multi-source data fusion, the method comprising: Acquire dialyzer data; perform region segmentation preprocessing on the dialyzer data to obtain preprocessed data of the dialyzer fiber bundle structure; The preprocessed data is parsed into a discrete feature dataset that characterizes the structural features. The discrete feature dataset is then processed to obtain the centroid coordinates of the core distribution area of ​​the dialyzer fiber bundle. An analysis plane is constructed based on the centroid coordinates, and a benchmark analysis framework that is compatible with the distribution characteristics of the dialyzer fiber bundle is established on the analysis plane. The benchmark analysis framework is divided into multiple levels to obtain multiple analysis sub-regions, and the discrete feature datasets are aggregated into the corresponding analysis sub-regions; based on the aggregation characteristics of the data within the analysis sub-regions, the corresponding regional feature correction factors are obtained. The aggregated feature data is corrected by a regional feature correction factor to identify coagulation regions in the feature data. The total area of ​​each analysis sub-region and the actual area of ​​the coagulation region in the feature data are calculated to obtain the calculation results. The coagulation region is quantified by combining the calculation results, and the coagulation feature quantification data of coagulation area ratio and coagulation point distribution density are extracted. The coagulation feature quantification data is input into a pre-trained intelligent coagulation grading processing model for analysis to obtain the coagulation grade determination result. The coagulation grade determination results are fused with the corresponding quantitative data of coagulation characteristics and patient identification information to obtain standardized coagulation assessment report data.

[0006] Further, dialyzer data is acquired; the dialyzer data is preprocessed by region segmentation to obtain preprocessed data of the dialyzer fiber bundle structure, including: The raw data of the dialyzer cross section is acquired by the acquisition device, and the raw data is dimensionally standardized and optically distorted to obtain standard data. Edge detection and contour extraction are performed on standard data to identify and segment the overall outer contour region of the dialyzer. Within the overall outer contour area, based on the known structural features of the dialyzer, the blood chamber region through which blood flows and the venous outlet region are identified and segmented; Within the blood chamber region, morphological processing and cluster analysis were used to separate regions of aggregated fiber bundle clusters and regions of discrete fiber bundle monomers. By fusing data from the fiber bundle cluster region and the fiber bundle individual region, preprocessed data for the complete spatial arrangement structure of the dialyzer fiber bundles is obtained.

[0007] Furthermore, the preprocessed data is parsed into discrete feature datasets characterizing structural features. These discrete feature datasets are then processed to obtain the centroid coordinates of the dialyzer fiber bundle core distribution region. An analysis plane is constructed based on these centroid coordinates, and a benchmark analysis framework adapted to the dialyzer fiber bundle distribution characteristics is established on this analysis plane, including: The pixel information representing the spatial arrangement structure of fiber bundles in the preprocessed data is parsed into a discrete feature dataset containing the geometric center coordinates, contour area and gray-scale mean information of each independent fiber bundle structure. Based on discrete feature datasets, the core distribution region of dialyzer fiber bundles is identified and delineated according to the contour area distribution and spatial density of each fiber bundle. The centroid coordinates of the core distribution area are obtained by calculating the weighted average of the geometric center coordinates of all fiber bundles within the core distribution area. A two-dimensional virtual analysis plane parallel to the physical cross-section of the dialyzer is established with the centroid coordinates as the origin; On a two-dimensional virtual analysis plane, a polar coordinate benchmark analysis framework covering the entire core distribution area is constructed based on the boundary contour of the core distribution area and the spatial distribution density field of the fiber bundles.

[0008] Furthermore, the benchmark analysis framework is divided into multiple levels to obtain several analysis sub-regions, and the discrete feature datasets are aggregated into the corresponding analysis sub-regions. Based on the aggregation characteristics of the data within the analysis sub-regions, corresponding regional feature correction factors are obtained, including: The baseline analysis framework is divided into concentric circles according to the preset partitioning strategy, resulting in multiple spatially continuous and non-overlapping analysis sub-regions. Match the geometric center coordinates of each fiber bundle in the discrete feature dataset with the spatial range of each analysis sub-region, and aggregate the feature data of each fiber bundle into the analysis sub-region where the geometric center of each fiber bundle is located; Collect the characteristic data of each analysis sub-region, and calculate the spatial distribution density and gray-scale distribution characteristics of fiber bundles within the analysis sub-region; Based on spatial distribution density and grayscale distribution characteristics, combined with preset background features, a regional feature correction factor is calculated to eliminate background differences and uneven distribution between regions.

[0009] Furthermore, the aggregated feature data is corrected using a regional feature correction factor to identify coagulation regions within the feature data. The total area of ​​each analytical sub-region and the actual area of ​​the coagulation region in the feature data are calculated to obtain the calculation results. The coagulation regions are quantified based on the calculation results, and the coagulation feature quantification data of the coagulation area ratio and coagulation point distribution density are extracted, including: The regional feature correction factor corresponding to each analysis sub-region is applied to the fiber bundle feature data collected in the analysis sub-region, and the gray mean in the fiber bundle feature data is normalized and corrected to obtain the corrected feature data with background homogenization. Based on the preset coagulation feature threshold, the corrected feature data is traversed and clustered to identify the data point set that meets the coagulation feature, and each data point set is mapped back to the original spatial location corresponding to each data point in order to determine the range of each coagulation region. Calculate the total physical area of ​​the sub-region where each coagulation region is located, and the actual projected area of ​​the coagulation region within the sub-region, to obtain the result data including the total area of ​​each region and the actual coagulation area. Based on the results data, the proportion of coagulation area in each analysis sub-region is calculated, and the number of coagulation points and the spatial dispersion of coagulation point distribution in the analysis sub-region are statistically analyzed to obtain quantitative data of coagulation characteristics that characterize the proportion of coagulation area and the density of coagulation point distribution.

[0010] Furthermore, the quantified coagulation feature data is input into a pre-trained intelligent coagulation grading processing model for analysis to obtain the coagulation grade determination results, including: Load the pre-trained coagulation grading intelligent processing model from the model library, and import the coagulation feature quantification dataset as input data into the loaded coagulation grading intelligent processing model. The intelligent coagulation grading processing model processes the imported input data, calculates the characteristic indicators of coagulation area ratio, coagulation point distribution density and spatial dispersion, and fuses the characteristic indicators based on preset rules to obtain comprehensive feature weights. Based on the comprehensive feature weights, the results are mapped to the preset hierarchical decision rules to obtain the corresponding coagulation level determination results.

[0011] Furthermore, the coagulation grade determination results are fused with the corresponding quantitative data of coagulation characteristics and patient identification information to obtain standardized coagulation assessment report data, including: Receive the coagulation feature quantification dataset corresponding to the coagulation grade determination result and the associated patient identification information; Based on the preset report structure rules, the coagulation grade determination results are matched and combined with the coagulation characteristic quantitative data to obtain structured intermediate data containing the grading conclusion and quantitative details; The structured intermediate data is associated and encapsulated with patient identification information, and a timestamp and a unique report identifier are added to obtain a complete coagulation assessment report; Based on the preset medical data transmission standards, the complete coagulation assessment report is processed to standardize the format, resulting in standard coagulation assessment report data that can be directly received and parsed by hemodialysis management information.

[0012] Secondly, the intelligent data processing system for dialyzer coagulation grading based on multi-source data fusion includes: The acquisition module is used to acquire dialyzer data; the dialyzer data is preprocessed by region segmentation to obtain preprocessed data of the dialyzer fiber bundle structure; The module is used to parse the preprocessed data into discrete feature datasets that characterize the structural features. The discrete feature datasets are processed to obtain the centroid coordinates of the core distribution area of ​​the dialyzer fiber bundle. The analysis plane is constructed based on the centroid coordinates, and a benchmark analysis framework adapted to the distribution characteristics of the dialyzer fiber bundle is established on the analysis plane. The partitioning module is used to divide the benchmark analysis framework into multiple levels, resulting in multiple analysis sub-regions, and to aggregate the discrete feature datasets into the corresponding analysis sub-regions; based on the aggregation characteristics of the data within the analysis sub-regions, the corresponding regional feature correction factors are obtained. The calculation module is used to correct the aggregated feature data through the regional feature correction factor, identify the coagulation region in the feature data, calculate the total area of ​​each analysis sub-region and the actual area of ​​the coagulation region in the feature data to obtain the calculation results; combine the calculation results to quantify the coagulation region and extract the coagulation feature quantification data of coagulation area ratio and coagulation point distribution density. The judgment module is used to input the quantified coagulation feature data into the pre-trained intelligent coagulation grading processing model for analysis and to obtain the coagulation grade judgment result. The processing module is used to fuse the coagulation grade determination results with the corresponding quantitative data of coagulation characteristics and patient identification information to obtain standardized coagulation assessment report data.

[0013] Thirdly, a computing device, comprising: One or more processors; A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0014] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0015] The above-described solution of the present invention has at least the following beneficial effects: Because it employs technical means such as dialyzer data region segmentation preprocessing, discrete feature analysis and polar coordinate benchmark analysis framework based on the centroid coordinates of the fiber bundle core distribution region, multi-level division of the benchmark framework concentric circles and calculation of regional feature correction factors, coagulation region correction identification and quantitative extraction of coagulation area ratio and coagulation point distribution density, analysis of pre-trained coagulation grading intelligent model and generation of standardized coagulation assessment reports, it overcomes the technical problems of strong subjectivity, vague grading standards, lack of objective quantitative data support, and difficulty in effectively linking assessment results with the hemodialysis management information system when using traditional visual assessment of dialyzer coagulation. This achieves objectivity, quantification, and automation of dialyzer coagulation grading, and the resulting standardized reports can be directly received and analyzed by the hemodialysis management information system, providing data basis for medical staff to accurately formulate and adjust anticoagulation protocols, ensuring the effectiveness of dialysis treatment and patient safety. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the intelligent data processing method for dialyzer coagulation grading based on multi-source data fusion provided in an embodiment of the present invention.

[0017] Figure 2 This is a schematic diagram of an intelligent data processing system for coagulation grading of dialyzers based on multi-source data fusion, provided by an embodiment of the present invention. Detailed Implementation

[0018] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0019] like Figure 1 As shown, embodiments of the present invention propose an intelligent data processing method for dialyzer coagulation grading based on multi-source data fusion, the method comprising the following steps: Step 1: Obtain dialyzer data; perform preprocessing on the dialyzer data by region segmentation to obtain preprocessed data of the dialyzer fiber bundle structure; Step 2: Parse the preprocessed data into a discrete feature dataset that characterizes the structural features. Process the discrete feature dataset to obtain the centroid coordinates of the core distribution area of ​​the dialyzer fiber bundle. Construct an analysis plane based on the centroid coordinates and establish a benchmark analysis framework on the analysis plane that is compatible with the distribution characteristics of the dialyzer fiber bundle. Step 3: Divide the benchmark analysis framework into multiple levels to obtain multiple analysis sub-regions, and aggregate the discrete feature datasets into the corresponding analysis sub-regions; based on the aggregation characteristics of the data within the analysis sub-regions, obtain the corresponding regional feature correction factors; Step 4: Correct the aggregated feature data using the regional feature correction factor, identify the coagulation region in the feature data, calculate the total area of ​​each analysis sub-region and the actual area of ​​the coagulation region in the feature data to obtain the calculation results; combine the calculation results to quantify the coagulation region, and extract the coagulation feature quantification data of coagulation area ratio and coagulation point distribution density. Step 5: Input the quantified coagulation feature data into the pre-trained intelligent coagulation grading processing model for analysis to obtain the coagulation grade determination result. Step 6: The coagulation grade determination results are fused with the corresponding quantitative data of coagulation characteristics and patient identification information to obtain standardized coagulation assessment report data.

[0020] In this embodiment of the invention, the technical means employed—including dialyzer data acquisition and region segmentation preprocessing, parsing preprocessed data into discrete feature datasets and constructing a benchmark analysis framework adapted to the distribution characteristics of fiber bundles using the centroid coordinates of the core distribution area of ​​the fiber bundles, multi-level division of the benchmark analysis framework and calculation of regional feature correction factors, identification of coagulation regions in the corrected feature data and quantitative extraction of coagulation area ratio and coagulation point distribution density, analysis and determination of coagulation level using a pre-trained intelligent model for coagulation grading, and fusion of grade results with coagulation quantification data and patient identification to obtain standardized assessment report data—overcome the technical problems of strong subjectivity, vague grading standards, lack of objective quantitative support, and difficulty in standardizing and integrating assessment results into medical information when using traditional visual assessment of dialyzer coagulation. This achieves objectivity, quantification, and automation of dialyzer coagulation grading, and the standardized report can be directly adapted to medical information, providing reliable data for medical staff to formulate and adjust anticoagulation protocols, thus ensuring the effectiveness of dialysis treatment and patient safety.

[0021] In a preferred embodiment of the present invention, step 1 above may include: Step 1.1: Acquire raw data of the dialyzer cross-section using a data acquisition device, and perform dimensional standardization and optical distortion correction on the raw data to obtain standard data. Specifically, this includes: using a high-resolution industrial camera as the data acquisition device, paired with an adjustable brightness uniform light source, horizontally fixing the dialyzer on a dedicated clamping platform, adjusting the camera height and angle to ensure the lens is perpendicular to the dialyzer cross-section, ensuring the dialyzer cross-section is fully presented in the field of view without obstruction, starting the device to acquire raw image data of the dialyzer cross-section, referring to the official physical dimension parameters provided by the dialyzer manufacturer, setting a unified mapping ratio between pixels and actual physical length, performing pixel-by-pixel dimensional standardization on the acquired raw data, so that the image data of dialyzers of different specifications or batches have a unified scale reference standard, and using optical distortion correction technology to calibrate the raw data region by region for distortion problems such as edge stretching and inconsistent sharpness between the center and edge caused by the optical characteristics of the camera lens, eliminating interference caused by shape distortion and proportion deviation, and finally obtaining standard data with uniform size, true shape, and no distortion interference.

[0022] Step 1.2 involves edge detection and contour extraction of the standard data to identify and segment the overall outer contour region of the dialyzer. Specifically, this includes: using gradient edge detection technology to scan the entire area of ​​the obtained standard data, capturing pixels in the image where grayscale values ​​change abruptly. These pixels constitute the boundary lines between the dialyzer and the background, as well as between different structures within the dialyzer. The detected boundary lines are connected to form a preliminary contour graphic. Then, noise filtering technology is used to remove discrete isolated pixels and small interference lines from the contour, retaining a continuous and complete effective contour. Based on the overall shape design features of the dialyzer, contour curves that can completely surround the main structure of the dialyzer are selected from all effective contours, and the region is delineated using the contour curves as boundaries.

[0023] Step 1.3: Within the overall outer contour area, based on the known structural features of the dialyzer, identify and segment the blood chamber region through which blood flows and the venous outlet region. Specifically, this includes: pre-entering the standard structural design information of the dialyzer, clarifying that the blood chamber region is the core functional area inside the dialyzer that houses the fiber bundle and through which blood circulates, with a regular cylindrical outline shape; and the venous outlet region, located at one end of the dialyzer, having a specific aperture and distribution pattern. Based on the known structural features, perform feature comparison and region scanning on the standard data within the segmented overall outer contour area to identify regions that conform to the outline shape and positional features of the blood chamber region, and simultaneously locate regions with the structural characteristics of the venous outlet region. Through contour segmentation technology, these two functional regions are separately marked and segmented, distinguishing the blood chamber region from the venous outlet region and avoiding cross-interference of data from different functional regions.

[0024] Step 1.4: Within the blood chamber region, morphological processing and cluster analysis are used to separate aggregated fiber bundle clusters and discrete fiber bundle individual regions. Specifically, within the segmented blood chamber region, morphological dilation is first performed, using structuring elements of a specific size to scan the image, filling in the small gaps and holes between fiber bundles to make the fiber bundle outlines more complete. Morphological erosion is then performed to restore the original shape and boundary details of the fiber bundles, while removing residual small noise points and impurities from the image. Spatial cluster analysis is then used to classify and identify all fiber bundles based on their spatial distribution and pixel density within the blood chamber region: fiber bundles that are spatially adjacent, have high pixel density, and form a continuous whole are classified as aggregated fiber bundle clusters; fiber bundles that are spatially independent, not connected to other fiber bundles, and have relatively dispersed pixel density are classified as discrete fiber bundle individual regions.

[0025] Step 1.5 involves fusing the data from the fiber bundle cluster region and the fiber bundle individual region to obtain preprocessed data for the complete spatial arrangement structure of the dialyzer fiber bundles. This includes automatically collecting complete data such as the spatial coordinates, contour morphology parameters, and pixel distribution range of the fiber bundle cluster region, while simultaneously summarizing the spatial location, contour size, and pixel grayscale values ​​of the corresponding fiber bundle individual region. Using data integration technology, the fiber bundle cluster region data and the fiber bundle individual region data are matched point-by-point and seamlessly stitched together according to the actual spatial relationship between the two types of regions within the blood chamber area, eliminating boundary gaps or data discontinuities generated during region separation. The stitched dataset is then optimized using a data fusion algorithm. This algorithm focuses on the synergistic complementarity and consistency calibration of the two types of region data, specifically achieved through four operations: data consistency verification, comparing the scale standards and parameter formats of the two types of region data to ensure that the spatial coordinates are based on the same coordinate system. The system and contour parameters are standardized to avoid data conflicts caused by differences in parameter settings during separation. Precise spatial positioning is achieved by fine-tuning minor positional shifts after splicing based on the overall structural characteristics of the blood chamber region, ensuring that the spatial connection between fiber bundle clusters and adjacent fiber bundle units conforms to the actual structural design of the dialyzer. Redundant data is intelligently removed by identifying and eliminating duplicate pixel information or parameter data recorded during splicing, reducing data redundancy and avoiding interference with analysis. Feature complementarity and integration are achieved by fusing and refining the aggregated distribution characteristics of the fiber bundle cluster region with the discrete distribution characteristics of the fiber bundle unit region, resulting in a comprehensive feature dataset that fully reflects the spatial relationship between the two types of fiber bundles. This dataset is then processed using a data fusion algorithm to ensure that the data completely and accurately reflects the actual arrangement of dialyzer fiber bundles within the blood chamber region. Finally, preprocessed data that comprehensively and accurately characterizes the complete spatial arrangement structure of the dialyzer fiber bundles provides data support for discrete feature analysis and coagulation region quantitative analysis.

[0026] In this embodiment of the invention, the technical approach of acquiring raw dialyzer cross-sectional data through a data acquisition device, first performing size standardization and optical distortion correction to obtain standard data, then performing edge detection and contour extraction on the standard data to segment the overall outer contour region of the dialyzer, and subsequently segmenting the blood chamber region and venous outlet region within the overall outer contour region based on structural features, and separating the fiber bundle cluster region and fiber bundle individual region within the blood chamber region through morphological processing and cluster analysis, and fusing the data of the two types of regions to obtain preprocessed data characterizing the complete spatial arrangement structure of the dialyzer fiber bundles, overcomes the technical problems in traditional dialyzer data preprocessing, such as inconsistent raw data size, optical distortion interference leading to data distortion, inability to accurately segment key functional regions, and difficulty in completely presenting the spatial distribution of fiber bundles, which in turn affects the quality of basic data for subsequent coagulation grading analysis. This achieves the technical effect of ensuring the accuracy of dialyzer coagulation grading analysis from the source by providing high-quality, high-reliability basic data for steps such as discrete feature dataset construction, benchmark analysis framework establishment, and coagulation region identification.

[0027] In a preferred embodiment of the present invention, step 2 above may include: Step 2.1 involves parsing the pixel information representing the spatial arrangement of fiber bundles in the preprocessed data into a discrete feature dataset containing the geometric center coordinates, contour area, and grayscale mean of each independent fiber bundle structure. Specifically, this includes: performing a full-area pixel-by-pixel scan analysis on the preprocessed data, focusing on the pixel information representing the spatial arrangement of fiber bundles; for each independent fiber bundle structure, locating its contour boundary using pixel coordinates, calculating the geometric center point of the contour boundary, determining the geometric center coordinates of the fiber bundle, counting the total number of pixels enclosed by the contour boundary of each independent fiber bundle, and calculating the contour area of ​​each independent fiber bundle by combining a preset pixel-to-actual area mapping ratio; extracting the grayscale values ​​of all pixels within each independent fiber bundle region, calculating the arithmetic mean of the grayscale values ​​to obtain the grayscale mean of each independent fiber bundle, and then organizing and summarizing the geometric center coordinates, contour area, and grayscale mean of each independent fiber bundle in a unified format to obtain a discrete feature dataset containing the key structural features of all independent fiber bundles.

[0028] Step 2.2: Based on the discrete feature dataset, the core distribution area of ​​the dialyzer fiber bundles is identified and delineated according to the contour area distribution and spatial density of each fiber bundle. Specifically, this includes: performing a comprehensive statistical analysis of the contour area of ​​all fiber bundles in the discrete feature dataset; determining the overall distribution range of the contour area through sorting and grouping; identifying the concentrated intervals of the contour area; clarifying the proportion of fiber bundles of different area levels in the overall fiber bundle; dividing several equally sized virtual spatial units according to the geometric center coordinates of each fiber bundle; counting the number of fiber bundles contained in each virtual spatial unit; calculating the number of fiber bundles distributed within a unit space, i.e., the spatial distribution density of the fiber bundles; obtaining a complete spatial density distribution map; and setting reasonable contour area thresholds and spatial density thresholds based on the high incidence of coagulation during clinical dialysis. The contour area threshold corresponds to the concentrated interval of the contour area, and the spatial density threshold is higher than the overall average spatial density. The area that simultaneously satisfies the condition that the contour area is within the concentrated interval and the spatial density is higher than the set threshold is designated as the core distribution area of ​​the dialyzer fiber bundles. The core distribution area is the key area where the fiber bundle distribution is most concentrated, which has the greatest impact on dialysis effect and the highest risk of coagulation.

[0029] Step 2.3: The centroid coordinates of the core distribution area are obtained by calculating the weighted average of the geometric center coordinates of all fiber bundles within the core distribution area. Specifically, this includes: extracting the geometric center coordinates and corresponding contour area data of all fiber bundles within the core distribution area from the discrete feature dataset; clarifying the spatial location and size of each fiber bundle; considering the different degrees of influence of fiber bundles of different sizes on the overall centroid of the core distribution area, the contour area of ​​each fiber bundle is directly used as the weight value of its geometric center coordinates. The larger the contour area, the higher the corresponding weight percentage, ensuring that the centroid calculation accurately reflects the spatial distribution of the core area. The geometric center coordinates of all fiber bundles are weighted and calculated separately. Multiplying the x-coordinate of each fiber bundle by its corresponding contour area yields a weighted value for the x-coordinate. The weighted values ​​of all fiber bundles' x-coordinates are summed to obtain a weighted total. Using the same method, the weighted total of the y-coordinates of all fiber bundles is calculated. The contour areas of all fiber bundles within the core distribution area are then calculated and summed to obtain the total weight. Dividing the weighted total of the x-coordinates by the total weight yields the weighted average of the x-coordinates of the geometric center of the core distribution area. Dividing the weighted total of the y-coordinates by the total weight yields the weighted average of the y-coordinates of the geometric center of the core distribution area. Combining these two weighted averages gives the centroid coordinates of the core distribution area, which is the spatial equilibrium center of all fiber bundles within the core area.

[0030] Step 2.4: Establish a two-dimensional virtual analysis plane parallel to the physical cross-section of the dialyzer, using the centroid coordinates as the origin. This includes: using the centroid coordinates as the absolute origin of the two-dimensional virtual analysis plane, with the absolute origin precisely corresponding to the spatial equilibrium center of the core distribution area; and determining the coordinate axis directions of the two-dimensional virtual analysis plane by referencing the actual spatial orientation of the dialyzer's physical cross-section. The horizontal coordinate axis is parallel to the transverse centerline of the dialyzer's physical cross-section, and the vertical coordinate axis is parallel to the longitudinal centerline of the dialyzer's physical cross-section. This setting ensures that the virtual analysis plane is completely consistent with the spatial orientation of the actual dialyzer structure, eliminating... To mitigate the positional discrepancy between virtual and real data, the scale units of the coordinate axes of the two-dimensional virtual analysis plane are set to maintain strict consistency with the pixel scale of the preprocessed data. For example, one pixel in the preprocessed data corresponds to 0.1 mm in the actual length of the dialyzer. The scale units of the virtual plane are also set proportionally. The precise mapping between the virtual plane coordinates and the actual physical dimensions of the dialyzer allows for comprehensive calibration of the entire coordinate system. By comparing the correspondence between the virtual coordinates and the actual position of the physical cross-section of the dialyzer multiple times, any coordinate position within the virtual analysis plane can be accurately mapped to the corresponding actual position of the physical cross-section of the dialyzer.

[0031] Step 2.5: On the two-dimensional virtual analysis plane, based on the boundary contour of the core distribution area and the spatial distribution density field of the fiber bundles, construct a polar coordinate benchmark analysis framework covering the entire core distribution area. Specifically, this includes: extracting the boundary contour data of the defined core distribution area on the established two-dimensional virtual analysis plane, clarifying the specific spatial range of the core distribution area on the virtual plane, and using this as the coverage boundary of the polar coordinate benchmark analysis framework. This ensures that the framework focuses only on the core fiber bundle region, excluding blank areas or non-core structures, thus improving analysis efficiency. A detailed analysis of the spatial distribution density field of the fiber bundles within the core distribution area is then performed, dividing the area into several virtual analysis sub-units of equal area. The number of fiber bundles within each virtual analysis sub-unit is counted, clarifying the density of fiber bundle distribution in different regions, and obtaining a complete spatial distribution density field map. The geocentric coordinates are used as the polar coordinates. The origin of the coordinate system is used to measure the maximum distance from the origin to the boundary contour of the core distribution area. This maximum distance is set as the radial range of the polar coordinates to ensure that the polar coordinate framework completely covers the entire core distribution area without missing any key fiber bundle components. Combined with the spatial distribution density field map, the radial and circumferential directions of the polar coordinates are differentiated. For areas with dense fiber bundle distribution, the radial and circumferential intervals are set more finely, for example, every 0.5 mm in the radial direction and every 10 degrees in the circumferential direction, to ensure the analysis accuracy of dense areas. For areas with sparse fiber bundle distribution, the intervals are appropriately widened, with every 1 mm in the radial direction and every 20 degrees in the circumferential direction, to ensure the analysis effect while avoiding resource waste. Through the above operations, a polar coordinate benchmark analysis framework adapted to the fiber bundle distribution characteristics is constructed.

[0032] In this embodiment of the invention, the pixel information representing the spatial arrangement of fiber bundles in the preprocessed data is parsed into a discrete feature dataset containing the geometric center coordinates, contour area, and grayscale mean of each independent fiber bundle. Based on this dataset, a core distribution region is identified and delineated according to the fiber bundle contour area distribution and spatial density. The weighted average of the geometric center coordinates of all fiber bundles within the core region is calculated to obtain the centroid coordinates. A two-dimensional virtual analysis plane parallel to the physical cross-section of the dialyzer is established with the centroid coordinates as the origin. Then, a polar coordinate benchmark analysis framework covering the entire core region is constructed by combining the boundary contour of the core distribution region and the spatial distribution density field of the fiber bundles. This technique overcomes the technical problems of traditional analysis, such as the lack of structured fiber bundle feature data support, ambiguous positioning of the core analysis area, and mismatch between the analysis framework and the actual distribution of fiber bundles, which leads to insufficient targeting and limited accuracy of subsequent coagulation assessment. It transforms the spatial arrangement characteristics of fiber bundles into discrete data that can be quantified and analyzed, accurately locates the core area of ​​coagulation assessment, and the constructed polar coordinate benchmark analysis framework can adapt to the distribution characteristics of fiber bundles and fully cover the core analysis area. It provides an accurate and suitable basic framework for sub-region division, coagulation area identification and quantitative analysis, and improves the targeting and accuracy of subsequent coagulation grading assessment.

[0033] In a preferred embodiment of the present invention, step 3 above may include: Step 3.1: Divide the baseline analysis framework into concentric circles according to the preset partitioning strategy to obtain multiple spatially continuous and non-overlapping analysis sub-regions. Specifically, this includes: referring to the constructed polar coordinate baseline analysis framework and combining the distribution law of the fiber bundle spatial distribution density field, formulating a preset concentric circle partitioning strategy. First, determine the number of radial partitioning layers. With the polar coordinate origin as the center, adjust the radial width of each layer according to the total radial range of the core distribution area and the density of fiber bundle distribution. The radial width of the inner layer area with dense fiber bundle distribution is set narrower, and the radial width of the outer layer area with sparse distribution is appropriately widened to ensure that the number of fiber bundles in each sub-region is within a reasonable range. Then, determine the number of circumferential partitioning segments. Divide evenly at 360 degrees so that each radial layer is divided into multiple equal fan-shaped areas. Perform a complete concentric circle partitioning operation on the polar coordinate baseline analysis framework in the order of first radial layering and then circumferential segmentation. Finally, obtain multiple spatially continuous and non-overlapping analysis sub-regions. Each analysis sub-region has a clear radial range and circumferential angle range.

[0034] Step 3.2 involves matching the geometric center coordinates of each fiber bundle in the discrete feature dataset with the spatial range of each analysis sub-region. The feature data of each fiber bundle is then aggregated into the analysis sub-region where its geometric center is located. Specifically, this includes: extracting the geometric center coordinates of each independent fiber bundle, along with all corresponding feature data such as contour area and grayscale mean, from the discrete feature dataset to ensure no data omissions; defining the spatial boundaries of each analysis sub-region, i.e., the radial distance interval and circumferential angle interval corresponding to each sub-region; and for each fiber bundle's geometric center coordinates, determining whether its radial distance falls within the radial interval of a certain analysis sub-region, and simultaneously determining whether its circumferential angle falls within the circumferential interval of that analysis sub-region. This dual determination achieves precise matching between coordinates and sub-regions. Once the analysis sub-region to which the geometric center coordinates of a fiber bundle belong is determined, all feature data corresponding to that fiber bundle are completely aggregated into the dedicated data storage unit of that analysis sub-region. This matching and aggregation operation is repeated until all fiber bundle feature data in the discrete feature dataset are accurately assigned to their corresponding analysis sub-regions. This achieves standardized aggregation of feature data according to spatial location, avoiding data chaos and providing orderly data support for regional statistical analysis.

[0035] Step 3.3 involves statistically analyzing the feature data collected for each sub-region, calculating the spatial distribution density and grayscale distribution characteristics of fiber bundles within the sub-region. Specifically, this includes: for each sub-region, retrieving all fiber bundle feature data collected in its dedicated data storage unit, counting the total number of fiber bundles contained within the sub-region, calculating the actual physical area of ​​the sub-region based on its radial width and circumferential angle, and combining the scale standard of the polar coordinate reference analysis framework; dividing the total number of fiber bundles within the sub-region by the actual physical area of ​​the sub-region to obtain the spatial distribution density of fiber bundles within the sub-region. This density value accurately reflects the density of fiber bundles within the region. The grayscale mean data of all fiber bundles within the sub-region is extracted, and the maximum, minimum, and arithmetic mean of the grayscale mean are statistically analyzed. The standard deviation of the grayscale mean is calculated. These indicators comprehensively characterize the grayscale distribution characteristics of fiber bundles within the sub-region, reflecting the overall level and uniformity of fiber bundle color intensity within the region, providing crucial feature references for determining coagulation areas.

[0036] Step 3.4: Based on spatial distribution density and grayscale distribution characteristics, combined with preset background characteristics, calculate the regional feature correction factor used to eliminate background differences and distribution inhomogeneity between regions. Specifically, this includes: statistically analyzing a large amount of dialyzer sample data in a non-coagulation state to obtain the standard spatial distribution density range and standard grayscale distribution characteristics of the fiber bundles under non-coagulation conditions. This standard data is set as the preset background characteristic and used as the benchmark for subsequent comparison and correction. For each analysis sub-region, the calculated actual spatial distribution density is compared with the standard spatial distribution density range in the preset background characteristics to calculate the density deviation value; the mean and standard deviation of the actual grayscale distribution characteristics are compared with the preset background characteristics... The standard grayscale mean and standard deviation in the background features are compared to calculate the grayscale deviation value. Based on the magnitude and direction of the density deviation value and grayscale deviation value, and combined with the correction experience summarized in clinical practice, a reasonable correction rule is formulated. That is, the larger the density deviation and grayscale deviation, the stronger the corresponding correction force. According to the correction rule, a unique regional feature correction factor is calculated for each analysis sub-region. The regional feature correction factor specifically offsets the influence of background differences and uneven fiber bundle distribution in that sub-region, ensuring that the feature data of all analysis sub-regions can be unified to the benchmark of the preset background features, eliminating inter-regional errors, and providing a consistent data foundation for cross-regional coagulation identification and quantitative analysis.

[0037] In this embodiment of the invention, by employing techniques such as dividing the polar coordinate benchmark analysis framework into concentric circles according to a preset strategy to obtain multiple spatially continuous and non-overlapping sub-regions, matching and aggregating the fiber bundle feature data in the discrete feature dataset to the corresponding sub-regions according to the geometric center coordinates, statistically analyzing the feature data of each sub-region and calculating the spatial distribution density and gray-scale distribution characteristics of the fiber bundles, and combining preset background features to calculate regional feature correction factors to eliminate background differences and uneven distribution between regions, the technical problems of coarse region division, disordered data aggregating, background differences between regions, and uneven fiber bundle distribution interfering with the accuracy of feature identification in traditional analysis are overcome. This achieves refined analysis region division and standardized feature data aggregating, effectively eliminating errors caused by background interference and uneven distribution in different regions. It provides a consistent and high-quality sub-region feature data foundation for subsequent accurate identification and quantitative analysis of coagulation regions, improving the accuracy and reliability of dialyzer coagulation grading assessment.

[0038] In a preferred embodiment of the present invention, step 4 above may include: Step 4.1: Apply the regional feature correction factors corresponding to each analysis sub-region to the fiber bundle feature data collected in the analysis sub-region, and normalize the gray-scale mean in the fiber bundle feature data to obtain corrected feature data with background homogenization. Specifically, this includes: extracting the regional feature correction factors corresponding to each analysis sub-region one by one to ensure that each correction factor corresponds precisely to its respective analysis sub-region without cross-contamination; for each analysis sub-region, retrieving all fiber bundle feature data collected in its dedicated data storage unit, focusing on the gray-scale mean parameter in the feature data, and performing targeted calculations on the regional feature correction factors of the analysis sub-region and the gray-scale mean of each fiber bundle to offset the gray-scale deviation of the fiber bundles in the sub-region caused by background differences and uneven distribution through the correction factors. For sub-regions with dense fiber bundle distribution and generally high gray level, the gray level is appropriately reduced by a correction factor; for sub-regions with sparse fiber bundle distribution and generally low gray level, the gray level is appropriately increased by a correction factor. After correction, the gray level of fiber bundles in all sub-regions is unified to the preset background baseline level, and finally, the corrected feature data with background homogenization is obtained.

[0039] Step 4.2: Based on the preset coagulation feature threshold, the corrected feature data is traversed and clustered to identify the data point set that meets the coagulation characteristics. Each data point set is then mapped back to its original spatial location to determine the range of each coagulation region. Specifically, this includes: based on a large amount of clinically labeled non-coagulating and coagulating dialyzer sample data, statistical analysis is used to determine the typical range of the average gray value of fiber bundles under coagulation conditions. Combined with clinical coagulation judgment criteria, a scientifically reasonable coagulation feature threshold is set. This threshold accurately distinguishes the gray value difference of fiber bundles between coagulation and non-coagulation states. The corrected feature data within each analysis sub-region is comprehensively traversed, and each data point is judged individually. If the average gray value of each fiber bundle is lower than the preset coagulation characteristic threshold, it is initially identified as a suspected coagulation data point. Cluster analysis is performed on all the initially identified suspected coagulation data points, and suspected coagulation data points that are spatially adjacent and have similar gray value characteristics are grouped into a data point set, resulting in multiple independent clusters of suspected coagulation data points. At the same time, isolated discrete suspected data points are removed to avoid noise interference. Each clustered data point set is back mapped to the original spatial position of the dialyzer physical cross section according to the corresponding geometric coordinate information. Combined with the spatial arrangement structure of the fiber bundle, the specific range of each data point set in the actual structure of the dialyzer is determined.

[0040] Step 4.3: Calculate the total physical area of ​​each coagulation region within its analysis sub-region, and the actual projected area of ​​the coagulation region within the analysis sub-region. This yields result data containing the total area of ​​each region and the actual coagulation area. Specifically, for each analysis sub-region, calculate the total physical area of ​​the analysis sub-region based on the radial range and circumferential angle in the polar coordinate reference analysis frame, combined with a preset pixel-to-actual physical area mapping ratio. Ensure that the area calculation result is completely consistent with the actual structural dimensions of the dialyzer. For each identified coagulation region within the analysis sub-region, count the number of all pixels contained within the boundary of the coagulation region, and then convert the number of pixels into the actual projected area of ​​the coagulation region according to the same pixel-to-actual physical area mapping ratio. If there are multiple independent coagulation regions within an analysis sub-region, calculate the actual projected area of ​​each coagulation region separately and then sum them to obtain the total actual projected area of ​​all coagulation regions within the analysis sub-region. Associate and store the total physical area of ​​each analysis sub-region with the corresponding total actual projected area of ​​the coagulation region to obtain result data containing both types of area data.

[0041] Step 4.4: Based on the results data, calculate the proportion of coagulation area within each analysis sub-region, and statistically analyze the number of coagulation points and the spatial dispersion of their distribution within the analysis sub-region to obtain quantitative data representing the coagulation characteristics of the proportion of coagulation area and the density of coagulation point distribution. Specifically, this includes: for each analysis sub-region, dividing the total actual projected area of ​​its corresponding coagulation area by the actual total physical area of ​​that analysis sub-region to obtain the proportion of coagulation area within the analysis sub-region. The proportion of coagulation area directly reflects the coverage of coagulation phenomena within the analysis sub-region. The number of independent coagulation areas within each analysis sub-region, or the number of fiber bundles within each coagulation area, is also counted. The quantity is considered as the number of coagulation points, and the total number of coagulation points in the analysis sub-region is obtained, reflecting the frequency of coagulation. The spatial distance between all coagulation points in each analysis sub-region is calculated, and the mean and standard deviation of the spatial distance are statistically analyzed to characterize the spatial dispersion of the coagulation point distribution. If the standard deviation is small, it indicates that the coagulation points are clustered, and if the standard deviation is large, it indicates that the coagulation points are dispersed. The coagulation area ratio, the total number of coagulation points, and the spatial dispersion of each analysis sub-region are integrated. Combined with the ratio of the total number of coagulation points to the actual total physical area of ​​the sub-region, the coagulation point distribution density is obtained. Finally, the quantitative data of coagulation characteristics including the coagulation area ratio and the coagulation point distribution density are obtained.

[0042] In this embodiment of the invention, the method involves applying the regional feature correction factor corresponding to each sub-region to the fiber bundle feature data of the corresponding sub-region, normalizing the gray-scale mean to obtain corrected feature data with background homogenization, performing traversal and cluster analysis on the corrected feature data based on a preset coagulation feature threshold, identifying data point sets that conform to coagulation features and mapping them back to their original spatial locations to determine the range of the coagulation area, calculating the total physical area of ​​each sub-region and the actual projected area of ​​the coagulation area within the sub-region to obtain result data, and calculating the proportion of the coagulation area in each sub-region and counting the number of coagulation points based on the result data. The technique of using dispersion to obtain quantitative data on coagulation characteristics overcomes the technical problems of traditional visual methods, such as the inability to quantify the proportion and distribution density of coagulation area, inaccurate coagulation identification due to background differences and uneven distribution between regions, and the lack of objective coagulation characteristic data to support grading judgment. It achieves more accurate coagulation area identification and quantification of coagulation characteristics, eliminates interference caused by regional background and distribution differences, obtains objective quantitative indicators that can be directly used for grading judgment, replaces subjective experience assessment, improves the consistency and accuracy of dialyzer coagulation grading, and provides reliable data basis for medical staff to formulate individualized anticoagulation plans.

[0043] In a preferred embodiment of the present invention, step 5 above may include: Step 5.1: Load the pre-trained intelligent coagulation grading model from the model library. Use the coagulation feature quantification dataset as input data and import it into the loaded intelligent coagulation grading model. Specifically, this includes: pre-building the model library. The intelligent coagulation grading model stored in the model library is a dedicated model built on a deep learning framework. Its core includes a feature integration module and a grading decision module. It is specifically designed and developed for dialyzer coagulation grading scenarios. The model has undergone multiple rounds of training and optimization with a large number of clinical samples. The training data comprehensively covers coagulation feature quantification data corresponding to different coagulation levels from 0 to 3, different mainstream dialyzer models on the market, and patient groups of different ages, genders, and underlying diseases. At the same time, it deeply integrates the fluctuation characteristics of pressure parameters such as arterial pressure, venous pressure, and transmembrane pressure from clinical diagnosis and treatment data as auxiliary training basis. Through iterative optimization and adjustment of the model's internal parameters, the model can accurately learn the mapping relationship between coagulation features and grading results in different scenarios.

[0044] The pre-trained coagulation grading intelligent processing model is retrieved from the model library via a model loading interface. During loading, model parameter verification and runtime environment adaptation are automatically completed: the parameter verification stage checks the integrity of the model's network structure, the effectiveness of training weights, and the consistency of feature processing logic to prevent abnormal operation due to corrupted parameters or version incompatibility; the runtime environment adaptation stage allocates computing power based on current hardware resources, matches the corresponding software dependency library versions, and extracts the coagulation feature quantification dataset. This dataset contains core quantitative indicators such as the proportion of coagulation area, coagulation point distribution density, and spatial dispersion of all analysis sub-regions, along with a summary of key diagnostic and treatment parameters during patient dialysis. These data collectively constitute the model's multi-dimensional input features, meeting the model's need for comprehensive information. To improve the accuracy of coagulation grading, the dataset is standardized according to its format. Following the input format requirements preset by the intelligent coagulation grading model, various quantitative indicators and summaries of diagnostic and treatment parameters are arranged and combined in an orderly manner, unifying the dimensional specifications and numerical ranges of the data. The system automatically verifies the completeness and rationality of the data, checking for missing data, values ​​exceeding normal physiological or physical ranges, and removing any identified abnormal or invalid data to prevent poor-quality data from affecting the model's analysis results. The standardized coagulation feature quantitative dataset is then imported into the already loaded intelligent coagulation grading model via a dedicated data transmission channel. During the import process, the data transmission progress and integrity are monitored in real time, and a data verification mechanism ensures that all quantitative indicators and summaries of diagnostic and treatment parameters are accurately transmitted to the model's feature integration module.

[0045] Step 5.2: The intelligent coagulation grading processing model processes the imported input data, calculates the feature indicators of coagulation area ratio, coagulation point distribution density, and spatial dispersion, and fuses the feature indicators based on preset rules to obtain comprehensive feature weights. Specifically, after receiving the input data, the intelligent coagulation grading processing model preprocesses the data, including data normalization and feature filtering. It maps the indicators such as coagulation area ratio, coagulation point distribution density, and spatial dispersion to the same data interval, and at the same time filters out the core features that have a significant impact on coagulation grading and removes redundant interference information. The model processes various quantitative indicators through its built-in feature calculation module, accurately calculating the average proportion of the coagulation area within the entire dialyzer, the proportion of the maximum coagulation area, and the standard deviation of the coagulation area proportion in each sub-region, comprehensively reflecting the coagulation coverage. It calculates the average number of coagulation points per unit area and the distribution proportion of coagulation points in different density intervals, clarifying the density of coagulation points. It also calculates the average distance and standard deviation of distance between all coagulation points, quantifying the spatial dispersion of coagulation points. Based on the core needs of clinical coagulation grading and statistical analysis results, it pre-sets feature fusion rules. The proportion of coagulation area directly correlates with the severity of coagulation, with the highest weight; the density of coagulation points reflects the coagulation diffusion trend, with the next highest weight; and spatial dispersion assists in determining the coagulation type, with a relatively low weight. Following the pre-set rules, the model fuses the three feature indicators through a weighted summation, comprehensively considering the influence of each indicator on coagulation grading, and ultimately obtaining a comprehensive feature weight that fully characterizes the coagulation status of the dialyzer.

[0046] Step 5.3: Based on the comprehensive feature weights, the data is mapped to the preset grading decision rules to obtain the corresponding coagulation grade determination results. Specifically, this includes: After receiving the input data, the intelligent coagulation grading processing model preprocesses the data, including data normalization and feature filtering. Indicators such as the proportion of coagulation area, the distribution density of coagulation points, and spatial dispersion are uniformly mapped to the same data interval. Simultaneously, core features that significantly affect coagulation grading are selected, and redundant interference information is removed. The model processes each quantitative indicator through its built-in feature calculation module, accurately calculating the average proportion of coagulation area, the maximum proportion of coagulation area, and the standard deviation of the proportion of coagulation area in each sub-region within the entire dialyzer range, comprehensively reflecting the coagulation coverage area; calculating the standard deviation of the proportion of coagulation area per unit area... The model calculates the average number of coagulation points and the distribution ratio of coagulation points in different density ranges to clarify the density of coagulation points. It also calculates the average distance and standard deviation between all coagulation points to quantify the spatial dispersion of coagulation points. Based on the core needs of clinical coagulation grading and statistical analysis results, a feature fusion rule is preset. Among these rules, the proportion of coagulation area directly relates to the severity of coagulation and has the highest weight, followed by the distribution density of coagulation points, which reflects the trend of coagulation diffusion. The spatial dispersion, which assists in judging the coagulation type, has a relatively low weight. The model fuses the three feature indicators according to this preset rule by weighted summation. It comprehensively considers the influence of each indicator on the coagulation grading and finally obtains the comprehensive feature weights that fully characterize the coagulation status of the dialyzer.

[0047] In this embodiment of the invention, a pre-trained intelligent coagulation grading processing model is loaded from a model library. The coagulation feature quantification dataset is used as input to the model. The model processes the input data to calculate the feature indicators of coagulation area ratio, coagulation point distribution density, and spatial dispersion. Based on preset rules, the feature indicators are fused to obtain a comprehensive feature weight. The comprehensive feature weight is then mapped to preset grading decision rules. Therefore, this method overcomes the technical problems of traditional visual methods, which rely on the subjective experience of medical staff, have vague grading standards, and cannot integrate multi-dimensional coagulation features for objective judgment, resulting in poor consistency and accuracy of coagulation grading results. This achieves intelligent and standardized processing of coagulation grading, realizes the scientific fusion and accurate analysis of multi-dimensional coagulation features, improves the consistency, accuracy, and reliability of coagulation grade determination results, and provides objective and authoritative decision-making basis for medical staff to quickly formulate individualized anticoagulation plans, thus preventing delays in treatment adjustment due to subjective judgment disputes.

[0048] In a preferred embodiment of the present invention, step 6 above may include: Step 6.1: Receive the coagulation feature quantification dataset and associated patient identification information corresponding to the coagulation grade determination result. Specifically, this includes: starting the data receiving module, monitoring the output data of the intelligent coagulation grading processing model in real time, and automatically receiving the coagulation grade determination result and the corresponding coagulation feature quantification dataset transmitted by the model after the model completes the coagulation grade determination. The coagulation feature quantification dataset contains core quantitative indicators such as the proportion of coagulation area, coagulation point distribution density, and spatial dispersion of coagulation points in all analysis sub-regions. At the same time, it receives the patient identification information associated with this dialysis assessment. The patient identification information includes unique identification information such as the patient's name, medical record number, and dialysis treatment number to ensure that each coagulation assessment result can be accurately matched with a specific patient. During the receiving process, the integrity of the data is automatically verified, and it is checked for problems such as missing coagulation grade determination results, incomplete quantitative indicators, or missing patient identification information. If data is found to be missing or abnormal, a data retransmission request is immediately sent to the model until complete and valid data is received.

[0049] Step 6.2: Based on the preset report structure rules, the coagulation grade determination results are matched and combined with the coagulation characteristic quantitative data to obtain structured intermediate data containing the grading conclusion and quantitative details. Specifically, this includes: pre-establishing preset report structure rules according to clinical treatment guidelines and medical report writing standards. These rules define the core modules that a coagulation assessment report should include, such as a grading conclusion module, a quantitative details module, and a feature analysis module. The grading conclusion module presents the final coagulation grade, the quantitative details module displays various coagulation characteristic quantitative indicators, and the feature analysis module briefly explains the clinical significance of key quantitative indicators. The preset report structure rules are then invoked to process the data. The collected coagulation grade determination results are precisely matched with the quantitative data of coagulation characteristics. The coagulation grade determination results are filled into the grading conclusion module. The average and maximum values ​​of the coagulation area ratio, the distribution of each sub-region, the specific values ​​and interval distribution of the coagulation point distribution density, and the statistical results of the spatial dispersion of coagulation points are filled into the quantitative details module according to the module classification. At the same time, the data are logically combined according to preset rules. The correspondence between the quantitative data and the coagulation grade is associated in the feature analysis module. For example, it describes the range of the coagulation area ratio and the distribution characteristics of the coagulation points corresponding to the coagulation grade. Finally, the structured intermediate data containing the grading conclusion and quantitative details is obtained in a logically clear and structurally standardized manner.

[0050] Step 6.3 involves associating and encapsulating the structured intermediate data with patient identification information, and adding a timestamp and a unique report identifier to obtain a complete coagulation assessment report. This includes: extracting patient identification information; associating and binding the patient identification information with the structured intermediate data to ensure that all patient assessment information corresponds one-to-one with individual identity information, avoiding confusion between patient information and assessment results; automatically obtaining the current time as a timestamp, accurate to the second, recording the generation time of the coagulation assessment report to provide a time basis for subsequent diagnosis and treatment data traceability; obtaining a unique report identifier, which uses a combination of letters and numbers and is globally unique, used to distinguish coagulation assessment reports from different patients and different dialysis sessions; and uniformly encapsulating the associated patient identification information, structured intermediate data, timestamp, and unique report identifier, integrating them into a complete coagulation assessment report according to a preset report format. The report content covers core content such as basic patient information, assessment time, coagulation grade, details of various quantitative indicators, and feature analysis.

[0051] Step 6.4: Based on the preset medical data transmission standards, the complete coagulation assessment report is standardized to obtain standard coagulation assessment report data that can be directly received and parsed by hemodialysis management information. This includes: pre-interfacing with commonly used clinical hemodialysis management information, clarifying supported data transmission formats, field definitions, encoding standards, and other requirements, integrating these requirements into the preset medical data transmission standards, and standardizing the complete coagulation assessment report according to these standards. First, the report's data structure is unified, and all content is rearranged according to the field names and order specified in the transmission standards. Then, the data is encoded to ensure that text information, numerical information, identifiers, etc., in the report conform to the encoding requirements of the transmission standards. Simultaneously, data format compatibility is verified to ensure that the date format, numerical precision, character length, etc., in the report can be correctly recognized by the hemodialysis management information. Through standardization, the complete coagulation assessment report is converted into standard coagulation assessment report data that can be directly received and parsed by hemodialysis management information. This includes standardized patient identification fields, assessment time fields, report unique identifier fields, coagulation grade fields, various coagulation characteristic quantitative indicator fields, and feature analysis fields.

[0052] In this embodiment of the invention, because it employs a method of receiving a coagulation feature quantification dataset corresponding to the coagulation grade determination result and associated patient identification information, and matching and combining the coagulation grade determination result with the coagulation feature quantification data based on preset report structure rules to obtain structured intermediate data containing grading conclusions and quantification details, and associating and encapsulating the structured intermediate data with patient identification information and adding a timestamp and a unique report identifier to obtain a complete coagulation assessment report, and then standardizing the format of the complete coagulation assessment report according to preset medical data transmission standards, it overcomes the shortcomings of traditional visual assessment methods, such as the lack of structured presentation of results and the inconsistency between patient information and coagulation assessment data. The technical issues of loose connections and inconsistent report formats preventing direct reception and analysis of hemodialysis management information, leading to poor data linkage and difficulties in standardized archiving and rapid reuse of assessment results, have been addressed by this solution. This solution achieves structured, standardized, and normalized coagulation assessment reports, enabling precise correlation between coagulation grading conclusions, quantitative data, and patient information. Report data can be directly received and analyzed by hemodialysis management information, improving the linkage and reusability of medical data. It provides efficient support for medical staff to quickly access assessment details, standardize the archiving of treatment data, and promptly develop individualized anticoagulation plans, avoiding delays in treatment adjustments due to incompatible data formats or chaotic data connections.

[0053] like Figure 2 As shown, embodiments of the present invention also provide an intelligent data processing system for dialyzer coagulation grading based on multi-source data fusion, comprising: The acquisition module is used to acquire dialyzer data; the dialyzer data is preprocessed by region segmentation to obtain preprocessed data of the dialyzer fiber bundle structure; The module is used to parse the preprocessed data into discrete feature datasets that characterize the structural features. The discrete feature datasets are processed to obtain the centroid coordinates of the core distribution area of ​​the dialyzer fiber bundle. The analysis plane is constructed based on the centroid coordinates, and a benchmark analysis framework adapted to the distribution characteristics of the dialyzer fiber bundle is established on the analysis plane. The partitioning module is used to divide the benchmark analysis framework into multiple levels, resulting in multiple analysis sub-regions, and to aggregate the discrete feature datasets into the corresponding analysis sub-regions; based on the aggregation characteristics of the data within the analysis sub-regions, the corresponding regional feature correction factors are obtained. The calculation module is used to correct the aggregated feature data through the regional feature correction factor, identify the coagulation region in the feature data, calculate the total area of ​​each analysis sub-region and the actual area of ​​the coagulation region in the feature data to obtain the calculation results; combine the calculation results to quantify the coagulation region and extract the coagulation feature quantification data of coagulation area ratio and coagulation point distribution density. The judgment module is used to input the quantified coagulation feature data into the pre-trained intelligent coagulation grading processing model for analysis and to obtain the coagulation grade judgment result. The processing module is used to fuse the coagulation grade determination results with the corresponding quantitative data of coagulation characteristics and patient identification information to obtain standardized coagulation assessment report data.

[0054] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A smart data processing method for dialyzer coagulation grading based on multi-source data fusion, characterized in that, The method includes: Acquire dialyzer data; perform region segmentation preprocessing on the dialyzer data to obtain preprocessed data of the dialyzer fiber bundle structure; The preprocessed data is parsed into a discrete feature dataset that characterizes the structural features. The discrete feature dataset is then processed to obtain the centroid coordinates of the core distribution area of ​​the dialyzer fiber bundle. An analysis plane is constructed based on the centroid coordinates, and a benchmark analysis framework that is compatible with the distribution characteristics of the dialyzer fiber bundle is established on the analysis plane. The benchmark analysis framework is divided into multiple levels to obtain multiple analysis sub-regions, and the discrete feature datasets are aggregated into the corresponding analysis sub-regions; based on the aggregation characteristics of the data within the analysis sub-regions, the corresponding regional feature correction factors are obtained. The aggregated feature data is corrected by a regional feature correction factor to identify coagulation regions in the feature data. The total area of ​​each analysis sub-region and the actual area of ​​the coagulation region in the feature data are calculated to obtain the calculation results. The coagulation region is quantified by combining the calculation results, and the coagulation feature quantification data of coagulation area ratio and coagulation point distribution density are extracted. The coagulation feature quantification data is input into a pre-trained intelligent coagulation grading processing model for analysis to obtain the coagulation grade determination result. The coagulation grade determination results are fused with the corresponding quantitative data of coagulation characteristics and patient identification information to obtain standardized coagulation assessment report data.

2. The intelligent data processing method for dialyzer coagulation grading based on multi-source data fusion according to claim 1, characterized in that, Acquire dialyzer data; Preprocessing the dialyzer data by region segmentation yields preprocessed data of the dialyzer fiber bundle structure, including: The raw data of the dialyzer cross section is acquired by the acquisition device, and the raw data is dimensionally standardized and optically distorted to obtain standard data. Edge detection and contour extraction are performed on standard data to identify and segment the overall outer contour region of the dialyzer. Within the overall outer contour area, based on the known structural features of the dialyzer, the blood chamber region through which blood flows and the venous outlet region are identified and segmented; Within the blood chamber region, morphological processing and cluster analysis were used to separate regions of aggregated fiber bundle clusters and regions of discrete fiber bundle monomers. By fusing data from the fiber bundle cluster region and the fiber bundle individual region, preprocessed data for the complete spatial arrangement structure of the dialyzer fiber bundles is obtained.

3. The intelligent data processing method for dialyzer coagulation grading based on multi-source data fusion according to claim 2, characterized in that, The preprocessed data is parsed into a discrete feature dataset representing structural characteristics. This discrete feature dataset is then processed to obtain the centroid coordinates of the core distribution region of the dialyzer fiber bundles. An analysis plane is constructed based on these centroid coordinates, and a benchmark analysis framework adapted to the distribution characteristics of the dialyzer fiber bundles is established on this plane, including: The pixel information representing the spatial arrangement structure of fiber bundles in the preprocessed data is parsed into a discrete feature dataset containing the geometric center coordinates, contour area and gray-scale mean information of each independent fiber bundle structure. Based on discrete feature datasets, the core distribution region of dialyzer fiber bundles is identified and delineated according to the contour area distribution and spatial density of each fiber bundle. The centroid coordinates of the core distribution area are obtained by calculating the weighted average of the geometric center coordinates of all fiber bundles within the core distribution area. A two-dimensional virtual analysis plane parallel to the physical cross-section of the dialyzer is established with the centroid coordinates as the origin; On a two-dimensional virtual analysis plane, a polar coordinate benchmark analysis framework covering the entire core distribution area is constructed based on the boundary contour of the core distribution area and the spatial distribution density field of the fiber bundles.

4. The intelligent data processing method for dialyzer coagulation grading based on multi-source data fusion according to claim 3, characterized in that, The benchmark analysis framework is divided into multiple levels to obtain multiple analysis sub-regions, and the discrete feature datasets are aggregated into the corresponding analysis sub-regions. Based on the clustering characteristics of data within the analyzed sub-regions, corresponding regional feature correction factors are obtained, including: The baseline analysis framework is divided into concentric circles according to the preset partitioning strategy, resulting in multiple spatially continuous and non-overlapping analysis sub-regions. Match the geometric center coordinates of each fiber bundle in the discrete feature dataset with the spatial range of each analysis sub-region, and aggregate the feature data of each fiber bundle into the analysis sub-region where the geometric center of each fiber bundle is located; Collect the characteristic data of each analysis sub-region, and calculate the spatial distribution density and gray-scale distribution characteristics of fiber bundles within the analysis sub-region; Based on spatial distribution density and grayscale distribution characteristics, combined with preset background features, a regional feature correction factor is calculated to eliminate background differences and uneven distribution between regions.

5. The intelligent data processing method for dialyzer coagulation grading based on multi-source data fusion according to claim 4, characterized in that, The aggregated feature data is corrected by a regional feature correction factor to identify coagulation regions in the feature data. The total area of ​​each sub-region and the actual area of ​​the coagulation region in the feature data are calculated to obtain the calculation results. The coagulation region is quantified based on the calculation results, and quantitative data on coagulation features, including the proportion of coagulation area and the distribution density of coagulation points, are extracted, including: The regional feature correction factor corresponding to each analysis sub-region is applied to the fiber bundle feature data collected in the analysis sub-region, and the gray mean in the fiber bundle feature data is normalized and corrected to obtain the corrected feature data with background homogenization. Based on the preset coagulation feature threshold, the corrected feature data is traversed and clustered to identify the data point set that meets the coagulation feature, and each data point set is mapped back to the original spatial location corresponding to each data point in order to determine the range of each coagulation region. Calculate the total physical area of ​​the sub-region where each coagulation region is located, and the actual projected area of ​​the coagulation region within the sub-region, to obtain the result data including the total area of ​​each region and the actual coagulation area. Based on the results data, the proportion of coagulation area in each analysis sub-region is calculated, and the number of coagulation points and the spatial dispersion of coagulation point distribution in the analysis sub-region are statistically analyzed to obtain quantitative data of coagulation characteristics that characterize the proportion of coagulation area and the density of coagulation point distribution.

6. The intelligent data processing method for dialyzer coagulation grading based on multi-source data fusion according to claim 5, characterized in that, The quantified coagulation feature data is input into a pre-trained intelligent coagulation grading processing model for analysis to obtain coagulation grade determination results, including: Load the pre-trained coagulation grading intelligent processing model from the model library, and import the coagulation feature quantification dataset as input data into the loaded coagulation grading intelligent processing model. The intelligent coagulation grading processing model processes the imported input data, calculates the characteristic indicators of coagulation area ratio, coagulation point distribution density and spatial dispersion, and fuses the characteristic indicators based on preset rules to obtain comprehensive feature weights. Based on the comprehensive feature weights, the results are mapped to the preset hierarchical decision rules to obtain the corresponding coagulation level determination results.

7. The intelligent data processing method for dialyzer coagulation grading based on multi-source data fusion according to claim 6, characterized in that, The coagulation grade determination results are fused with corresponding quantitative data on coagulation characteristics and patient identification information to obtain standardized coagulation assessment report data, including: Receive the coagulation feature quantification dataset corresponding to the coagulation grade determination result and the associated patient identification information; Based on the preset report structure rules, the coagulation grade determination results are matched and combined with the coagulation characteristic quantitative data to obtain structured intermediate data containing the grading conclusion and quantitative details; The structured intermediate data is associated and encapsulated with patient identification information, and a timestamp and a unique report identifier are added to obtain a complete coagulation assessment report; Based on the preset medical data transmission standards, the complete coagulation assessment report is processed to standardize the format, resulting in standard coagulation assessment report data that can be directly received and parsed by hemodialysis management information.

8. A smart data processing system for dialyzer coagulation grading based on multi-source data fusion, wherein the system implements the method as described in any one of claims 1 to 7, characterized in that, include: The acquisition module is used to acquire dialyzer data; Preprocessing the dialyzer data by region segmentation yields preprocessed data of the dialyzer fiber bundle structure. The module is used to parse the preprocessed data into discrete feature datasets that characterize the structural features. The discrete feature datasets are processed to obtain the centroid coordinates of the core distribution area of ​​the dialyzer fiber bundle. The analysis plane is constructed based on the centroid coordinates, and a benchmark analysis framework adapted to the distribution characteristics of the dialyzer fiber bundle is established on the analysis plane. The partitioning module is used to divide the benchmark analysis framework into multiple levels, resulting in multiple analysis sub-regions, and to aggregate the discrete feature datasets into the corresponding analysis sub-regions; based on the aggregation characteristics of the data within the analysis sub-regions, the corresponding regional feature correction factors are obtained. The calculation module is used to correct the aggregated feature data through the regional feature correction factor, identify the coagulation region in the feature data, calculate the total area of ​​each analysis sub-region and the actual area of ​​the coagulation region in the feature data to obtain the calculation results; combine the calculation results to quantify the coagulation region and extract the coagulation feature quantification data of coagulation area ratio and coagulation point distribution density. The judgment module is used to input the quantified coagulation feature data into the pre-trained intelligent coagulation grading processing model for analysis and to obtain the coagulation grade judgment result. The processing module is used to fuse the coagulation grade determination results with the corresponding quantitative data of coagulation characteristics and patient identification information to obtain standardized coagulation assessment report data.

9. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images