Corneal endothelial function prediction and diagnosis system

Abstract

The invention discloses a corneal endothelial function prediction and diagnosis system. The system comprises a data storage and management module which is responsible for storing collected cell space coordinate data and a JSON format file generated by subsequent analysis; import, export, version management and historical operation record storage of data are supported, and data integrity and traceability are ensured; the Voronoi diagram generation module is used for generating a corresponding Voronoi space division diagram based on the input coordinate data of the cell center point through a built-in Voronoi diagram algorithm; the Voronoi diagram generation module supports batch processing of multiple groups of data; the invention relates to the technical field of medical diagnosis. According to the corneal endothelial function prediction and diagnosis system, corneal endothelial form generation is carried out based on a gravity center iterative algorithm, a Json file of a generated image is further analyzed to complete corneal endothelial function evaluation, and a new thought is provided for diagnosis of corneal endothelial diseases.

Description

Technical Field

[0001] This invention relates to the field of medical diagnostic technology, specifically to a corneal endothelial function prediction and diagnostic system. Background Technology

[0002] There are approximately 12.7 million people worldwide suffering from corneal blindness. Corneal endothelial disease is the leading cause of corneal blindness. Current clinical treatment strategies rely on late-stage corneal transplantation, and the lack of early diagnostic technology severely restricts the effectiveness of disease intervention. Therefore, to overcome this technological bottleneck, a corneal endothelial cell assessment and prediction system based on a centroid iterative algorithm is constructed.

[0003] This system is applicable to the assessment and prediction of endothelial function in various corneal endothelial diseases, including those following corneal transplantation and before and after cataract surgery. It can effectively advance the treatment window, improve the overall diagnosis and treatment level of corneal endothelial diseases, enhance the patient's medical experience, promote the intelligent development of ophthalmic clinical diagnosis and treatment, and provide an innovative model for building an intelligent eye disease diagnosis and treatment system. Summary of the Invention

[0004] To achieve the above objectives, the present invention provides the following technical solution: a corneal endothelial function prediction and diagnostic system, comprising:

[0005] Data storage and management module: responsible for storing the collected cell spatial coordinate data and JSON format files generated by subsequent analysis; supports data import, export, version management, and saving of historical operation records to ensure data integrity and traceability;

[0006] Voronoi diagram generation module: Based on the input cell center point coordinate data, it automatically generates the corresponding Voronoi spatial partitioning diagram using the built-in Voronoi diagram algorithm; this Voronoi diagram generation module supports batch processing of multiple sets of data to ensure the accuracy and stability of the generated graph;

[0007] Cell score calculation module: Based on Voronoi polygon vertex data, it calculates the morphological score of each cell; the calculation method combines the polygon area and the ratio of the diagonal length of non-adjacent vertices to quantify the cell size and shape regularity; this cell score calculation module outputs cell score data for subsequent visualization and analysis.

[0008] Data interaction and editing module: Provides a user interface that supports dynamic editing of cell data, including functions such as deleting specified cells, undoing operations, and resetting data; supports users to input cell numbers for data filtering and operations, ensuring the flexibility and accuracy of analysis;

[0009] The visualization module is responsible for visually displaying Voronoi diagrams and cell scores on the user interface using color mapping. It supports interactive operations such as score display / hiding switching, graph zooming, panning, and screenshotting, enhancing the user experience. It provides multiple color mapping schemes and font size adjustments to meet the needs of different users.

[0010] Multi-group data similarity comparison module: Supports importing multiple groups of corneal endothelial cell data, and calculates the similarity between data groups based on multi-dimensional indicators such as cell spatial distribution, morphological characteristics, number ratio, density and spatial structure; adopts a weighted comprehensive algorithm, supports user-defined weight parameters, and achieves flexible and accurate similarity assessment; provides numerical and graphical display of similarity results to assist in clinical diagnosis and scientific research analysis.

[0011] A method for predicting and diagnosing corneal endothelial function includes the following steps:

[0012] Step A1: Data Acquisition and Preprocessing;

[0013] Image data of corneal endothelial cells is acquired, typically using high-resolution microscopic images. Image processing algorithms are used to preprocess the raw images to extract the spatial coordinates of the cell center points; the extracted cell location data is then converted into point cloud data and stored as a JSON-formatted point set data file.

[0014] Step A2: Cell distribution analysis and Voronoi diagram construction;

[0015] The import function reads the corneal endothelial cell data file in JSON format selected by the user.

[0016] Parse the JSON file and extract the cell IDs and their corresponding two-dimensional coordinate data.

[0017] Using the extracted coordinate data, the built-in Voronoi diagram algorithm module is invoked to generate the corresponding Voronoi diagram;

[0018] The generated Voronoi diagram is visualized in the core area of ​​the software interface, with cell regions distinguished by different colors, and the color mapping is based on cell fractions.

[0019] Step A3: Calculate cell fractions from Voronoi polygon vertex data

[0020] Based on the generated Voronoi polygon data, the morphological features of the Voronoi region corresponding to each cell are quantified, and the cell score is calculated to reflect the size and shape regularity of the cell.

[0021] The formula for calculating the cell fraction is as follows:

[0022]

[0023] Where C is a constant, S is the area of ​​the polygon, and R is the ratio of the maximum diagonal length to the minimum diagonal length of the non-adjacent vertices of the polygon;

[0024] The formula for the area S of the polygon is as follows:

[0025]

[0026] Wherein, the length D of the maximum diagonal of the non-adjacent vertices of the polygon max and minimum diagonal length D min The ratio R is as follows:

[0027]

[0028] Step A4: Interactive Operations and Data Management;

[0029] The interactive operations and data management specifically include:

[0030] Data Deletion: Users can enter a specific cell number (e.g., highest score) on the interface. The system parses the input and deletes the corresponding cell data from the model. After deletion, the system automatically updates.

[0031] Voronoi diagram, recalculate cell fractions, and display the spatial distribution of remaining cells in the interface with different colors or markers;

[0032] Undo Operation: Before each data change, the system automatically saves the current state to the operation history stack. The current state includes all cell positions, Voronoi polygons, scores, and interface display parameters. When the user clicks the "Undo" button, the system pops the last saved state from the operation history stack and restores it to that state, ensuring that the user can safely undo erroneous operations or adjust analysis parameters.

[0033] Reset Operation: Users can select the "Reset" button. The system will clear all current analysis results and operation history, reload the original data (i.e., the initial state of the imported JSON file), and restore the model to its initial state. This operation is suitable when users want to restart the analysis process.

[0034] Data import: Users can select corneal endothelial cell data files in JSON format through the "Import File" function; the system parses the file content, extracts information such as cell number, location and morphological characteristics, automatically generates Voronoi diagram and cell score, and updates the interface display;

[0035] Data Export: After the current analysis is completed, users can choose to "export JSON" or "export image (PNG / TIFF)" to save information such as the spatial distribution, cell number, score, and Voronoi polygons of the current model to a file for subsequent analysis or reporting.

[0036] Step A5: Compare the similarity of multiple sets of data;

[0037] It supports importing multiple sets of corneal endothelial cell spatial distribution data in JSON format and uses multiple indicators to perform multi-dimensional similarity evaluation.

[0038] Preferably, step A5 involves multi-dimensional similarity evaluation of multiple indicators, including the following steps:

[0039] Step Q1: KDE spatial distribution similarity, structural feature similarity, morphological feature similarity, graph theory feature similarity, and density similarity;

[0040] Among them, KDE spatial distribution similarity uses Gaussian kernel density estimation to model the spatial distribution of cell points, and then measures the consistency of the overall spatial distribution by calculating the similarity between the two distributions.

[0041] Structural feature similarity is achieved by constructing a Voronoi diagram and analyzing the number of neighbors for each cell, reflecting the local structural features of the tissue.

[0042] Morphological features: Extract the geometric features of each cell polygon to measure the similarity of cell shapes;

[0043] Graph-based feature similarity: Modeling the cellular space as a graph and extracting its structural features to measure the overall connectivity of the tissue;

[0044] Density similarity: It reflects the uniformity of cell distribution by comparing the cell density in the effective region of two samples;

[0045] Step Q2: Calculate the overall similarity by summing the similarities of the five indicators using preset or adjusted weights. The specific similarity calculation formula is as follows:

[0046] Similarity Total =ω KDE ×Similarity KDE +ω Structure ×Similarity Structure

[0047] +ω Morphology ×Similairty Morphology +ω Graph ×Similairty Graphy +ωDensity ×Similairty Density .

[0048] Preferably, the KDE spatial distribution similarity calculation steps are as follows:

[0049] Step S1: Using the spatial points of each group of cells as samples, calculate the density value on the defined grid using nuclear density estimation;

[0050]

[0051] Where x represents a point in space, x i Let represent the i-th sample point, n represent the total number of sample points, h represent the bandwidth parameter used to control the smoothness of the kernel function, and K represent the kernel function.

[0052] Step S2: Calculate the density distributions p(x,y) and q(x,y) of the two samples;

[0053] Step S3: Use JS divergence to measure the difference between two density distributions:

[0054]

[0055] in, D KL The Kullback-Leibler divergence;

[0056] Step S4, similar meaning is:

[0057] Similarity KDE =1-JSD(p,q);

[0058] Structural feature similarity: By constructing a Voronoi diagram, the number of neighbors of each cell, i.e. the number of polygon vertices, is analyzed to reflect the local structural features of the tissue.

[0059] Preferably, the structural feature similarity calculation step specifically includes:

[0060] Step D1: Using the Voronoi diagram, count the number of neighbors of each cell, i.e., the number of vertices of the polygon;

[0061] Step D2: Construct a histogram of the number of neighbors, set the maximum number of neighbors to 10, and normalize it into probability distribution vectors h1 and h2;

[0062] Step D3: Calculate the cosine similarity between the two histograms:

[0063]

[0064] Preferably, the morphological feature calculation steps are as follows:

[0065] Step F1: Calculate the area, perimeter, circularity, aspect ratio, and Hu moment of each polygon;

[0066] Step F2: Normalize each feature;

[0067] Step F3: Use Wasserstein distance (f1, f2) to measure the difference between the two feature distributions;

[0068] Step F4: Convert to similarity:

[0069]

[0070] Preferably, the graph theory feature similarity calculation step specifically includes:

[0071] Step G1: Construct the graph: Nodes are cells, and edges are adjacency relationships, i.e., Delaunay triangulation;

[0072] Step G2: Extract features: degree distribution, clustering coefficient distribution, shortest path length distribution, and global statistics;

[0073] Step G3: Calculate the L1 distance for each distribution:

[0074]

[0075] Where, p i and q i Let n represent the component values ​​of the two vectors in the i-th dimension, where n is the dimension of the vector;

[0076] Step G4: Combine global statistics to form the overall graph structure similarity:

[0077]

[0078] Among them, D i ω represents the difference measure of the structural features of the i-th graph. i This represents the weight coefficient of the corresponding feature.

[0079] Preferably, the density similarity calculation step specifically includes:

[0080] Step H1: Count the number of cell points in the effective region of each sample group;

[0081] Step H2, Calculate density:

[0082]

[0083] Where Num represents the number of cells in the effective region, and S represents the area of ​​the effective region.

[0084] Step H3: Compare the ratio of the two densities:

[0085]

[0086] This invention provides a corneal endothelial function prediction and diagnostic system. It has the following beneficial effects:

[0087] (I) This corneal endothelial function prediction and diagnosis system, based on the automatic spatial division and cell fraction calculation of the Voronoi diagram, greatly improves the efficiency of corneal endothelial cell morphology analysis and avoids the inefficiency and subjectivity of traditional manual observation.

[0088] (ii) This corneal endothelial function prediction and diagnostic system achieves objective quantification of cell morphology by using a fractional calculation method that combines area and shape ratio, thereby improving the accuracy of cell status assessment.

[0089] (III) This corneal endothelial function prediction and diagnostic system uses color mapping technology to present complex cell morphology information intuitively, making it easier for clinicians and researchers to quickly identify abnormal cells and lesion areas.

[0090] (iv) This corneal endothelial function prediction and diagnostic system supports dynamic editing, undoing, resetting, and importing and exporting of cell data in multiple formats, thus meeting diverse clinical and research needs.

[0091] (v) This corneal endothelial function prediction and diagnostic system, through multi-indicator comprehensive evaluation of the similarity between different data groups, supports disease progression monitoring and treatment effect evaluation, thereby enhancing the practical value of the system. Attached Figure Description

[0092] Figure 1 This is a schematic diagram illustrating the generation of Voronoi based on a JOSN file according to the present invention;

[0093] Figure 2 This is a schematic diagram of the cell fraction display of the present invention;

[0094] Figure 3 This is a schematic diagram illustrating the similarity comparison of the present invention. Detailed Implementation

[0095] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0096] Example 1, please refer to Figure 1-3 The present invention provides a technical solution:

[0097] A corneal endothelial function prediction and diagnostic system, comprising:

[0098] Data storage and management module: responsible for storing the collected cell spatial coordinate data and JSON format files generated by subsequent analysis; supports data import, export, version management, and saving of historical operation records to ensure data integrity and traceability;

[0099] Voronoi diagram generation module: Based on the input cell center point coordinate data, it automatically generates the corresponding Voronoi spatial partitioning diagram using the built-in Voronoi diagram algorithm; this Voronoi diagram generation module supports batch processing of multiple sets of data to ensure the accuracy and stability of the generated graph;

[0100] Cell score calculation module: Based on Voronoi polygon vertex data, it calculates the morphological score of each cell; the calculation method combines the polygon area and the ratio of the diagonal length of non-adjacent vertices to quantify the cell size and shape regularity; this cell score calculation module outputs cell score data for subsequent visualization and analysis.

[0101] Data interaction and editing module: Provides a user interface that supports dynamic editing of cell data, including functions such as deleting specified cells, undoing operations, and resetting data; supports users to input cell numbers for data filtering and operations, ensuring the flexibility and accuracy of analysis;

[0102] The visualization module is responsible for visually displaying Voronoi diagrams and cell scores on the user interface using color mapping. It supports interactive operations such as score display / hiding switching, graph zooming, panning, and screenshotting, enhancing the user experience. It provides multiple color mapping schemes and font size adjustments to meet the needs of different users.

[0103] Multi-group data similarity comparison module: Supports importing multiple groups of corneal endothelial cell data, and calculates the similarity between data groups based on multi-dimensional indicators such as cell spatial distribution, morphological characteristics, number ratio, density and spatial structure; adopts a weighted comprehensive algorithm, supports user-defined weight parameters, and achieves flexible and accurate similarity assessment; provides numerical and graphical display of similarity results to assist in clinical diagnosis and scientific research analysis.

[0104] Example 2: Based on Example 1, the present invention provides a technical solution:

[0105] A method for predicting and diagnosing corneal endothelial function includes the following steps:

[0106] Step A1: Data Acquisition and Preprocessing;

[0107] Image data of corneal endothelial cells is acquired, typically using high-resolution microscopic images. Image processing algorithms are used to preprocess the raw images to extract the spatial coordinates of the cell center points; the extracted cell location data is then converted into point cloud data and stored as a JSON-formatted point set data file.

[0108] Step A2: Cell distribution analysis and Voronoi diagram construction;

[0109] The import function reads the corneal endothelial cell data file in JSON format selected by the user.

[0110] Parse the JSON file and extract the cell IDs and their corresponding two-dimensional coordinate data.

[0111] Using the extracted coordinate data, the built-in Voronoi diagram algorithm module is invoked to generate the corresponding Voronoi diagram;

[0112] The generated Voronoi diagram is visualized in the core area of ​​the software interface, with cell regions distinguished by different colors, and the color mapping is based on cell fractions.

[0113] Step A3: Calculate cell fractions from Voronoi polygon vertex data

[0114] Based on the generated Voronoi polygon data, the morphological features of the Voronoi region corresponding to each cell are quantified, and the cell score is calculated to reflect the size and shape regularity of the cell.

[0115] The formula for calculating the cell fraction is as follows:

[0116]

[0117] Where C is a constant, S is the area of ​​the polygon, and R is the ratio of the maximum diagonal length to the minimum diagonal length of the non-adjacent vertices of the polygon;

[0118] The formula for the area S of the polygon is as follows:

[0119]

[0120] Wherein, the length D of the maximum diagonal of the non-adjacent vertices of the polygon max and minimum diagonal length D min The ratio R is as follows:

[0121]

[0122] Step A4: Interactive Operations and Data Management;

[0123] The interactive operations and data management specifically include:

[0124] Data Deletion: Users can enter a specific cell number (e.g., highest score) on the interface. The system parses the input and deletes the corresponding cell data from the model. After deletion, the system automatically updates.

[0125] Voronoi diagram, recalculate cell fractions, and display the spatial distribution of remaining cells in the interface with different colors or markers;

[0126] Undo Operation: Before each data change, the system automatically saves the current state to the operation history stack. The current state includes all cell positions, Voronoi polygons, scores, and interface display parameters. When the user clicks the "Undo" button, the system pops the last saved state from the operation history stack and restores it to that state, ensuring that the user can safely undo erroneous operations or adjust analysis parameters.

[0127] Reset Operation: Users can select the "Reset" button. The system will clear all current analysis results and operation history, reload the original data (i.e., the initial state of the imported JSON file), and restore the model to its initial state. This operation is suitable when users want to restart the analysis process.

[0128] Data import: Users can select corneal endothelial cell data files in JSON format through the "Import File" function; the system parses the file content, extracts information such as cell number, location and morphological characteristics, automatically generates Voronoi diagram and cell score, and updates the interface display;

[0129] Data Export: After the current analysis is completed, users can choose to "export JSON" or "export image (PNG / TIFF)" to save information such as the spatial distribution, cell number, score, and Voronoi polygons of the current model to a file for subsequent analysis or reporting.

[0130] Step A5: Compare the similarity of multiple sets of data;

[0131] It supports importing multiple sets of corneal endothelial cell spatial distribution data in JSON format and uses multiple indicators to perform multi-dimensional similarity evaluation.

[0132] Example 3: Based on Example 2, the present invention provides a technical solution:

[0133] Step A5 involves a multi-dimensional similarity evaluation using multiple indicators, including the following steps:

[0134] Step Q1: KDE spatial distribution similarity, structural feature similarity, morphological feature similarity, graph theory feature similarity, and density similarity;

[0135] Among them, KDE spatial distribution similarity uses Gaussian kernel density estimation to model the spatial distribution of cell points, and then measures the consistency of the overall spatial distribution by calculating the similarity between the two distributions.

[0136] Structural feature similarity is achieved by constructing a Voronoi diagram and analyzing the number of neighbors for each cell, reflecting the local structural features of the tissue.

[0137] Morphological features: Extract the geometric features of each cell polygon to measure the similarity of cell shapes;

[0138] Graph-based feature similarity: Modeling the cellular space as a graph and extracting its structural features to measure the overall connectivity of the tissue;

[0139] Density similarity: It reflects the uniformity of cell distribution by comparing the cell density in the effective region of two samples;

[0140] Step Q2: Calculate the overall similarity by summing the similarities of the five indicators using preset or adjusted weights. The specific similarity calculation formula is as follows:

[0141] Similarity Total =ω KDE ×Similarity KDE +ω Structure ×Similarity Structure

[0142] +ω Morphology ×Similairty Morphology +ω Graph ×Similairty Graphy +ω Density ×Similairty Density .

[0143] The specific steps for calculating the spatial distribution similarity of KDE are as follows:

[0144] Step S1: Using the spatial points of each group of cells as samples, calculate the density value on the defined grid using nuclear density estimation;

[0145]

[0146] Where x represents a point in space, x i Let represent the i-th sample point, n represent the total number of sample points, h represent the bandwidth parameter used to control the smoothness of the kernel function, and K represent the kernel function.

[0147] Step S2: Calculate the density distributions p(x,y) and q(x,y) of the two samples;

[0148] Step S3: Use JS divergence to measure the difference between two density distributions:

[0149]

[0150] in, D KL The Kullback-Leibler divergence;

[0151] Step S4, similar meaning is:

[0152] Similarity KDE =1-JSD(p,q);

[0153] Structural feature similarity: By constructing a Voronoi diagram, the number of neighbors of each cell, i.e. the number of polygon vertices, is analyzed to reflect the local structural features of the tissue.

[0154] The specific steps for calculating the structural feature similarity are as follows:

[0155] Step D1: Using the Voronoi diagram, count the number of neighbors of each cell, i.e., the number of vertices of the polygon;

[0156] Step D2: Construct a histogram of the number of neighbors, set the maximum number of neighbors to 10, and normalize it into probability distribution vectors h1 and h2;

[0157] Step D3: Calculate the cosine similarity between the two histograms:

[0158]

[0159] The specific steps for calculating the morphological features are as follows:

[0160] Step F1: Calculate the area, perimeter, circularity, aspect ratio, and Hu moment of each polygon;

[0161] Step F2: Normalize each feature;

[0162] Step F3: Use Wasserstein distance (f1, f2) to measure the difference between the two feature distributions;

[0163] Step F4: Convert to similarity:

[0164]

[0165] The specific steps for calculating the graph theory feature similarity are as follows:

[0166] Step G1: Construct the graph: Nodes are cells, and edges are adjacency relationships, i.e., Delaunay triangulation;

[0167] Step G2: Extract features: degree distribution, clustering coefficient distribution, shortest path length distribution, and global statistics;

[0168] Step G3: Calculate the L1 distance for each distribution:

[0169]

[0170] Where, p i and q i Let n represent the component values ​​of the two vectors in the i-th dimension, where n is the dimension of the vector;

[0171] Step G4: Combine global statistics to form the overall graph structure similarity:

[0172]

[0173] Among them, D i ω represents the difference measure of the structural features of the i-th graph. i This represents the weight coefficient of the corresponding feature.

[0174] The density similarity calculation steps are as follows:

[0175] Step H1: Count the number of cell points in the effective region of each sample group;

[0176] Step H2, Calculate density:

[0177]

[0178] Where Num represents the number of cells in the effective region, and S represents the area of ​​the effective region.

[0179] Step H3: Compare the ratio of the two densities:

[0180]

[0181] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0182] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A corneal endothelial function predictive diagnosis system, characterized by, Comprise: Data storage and management module: responsible for storing the collected cell spatial coordinate data and the subsequent analysis generated JSON format file; Support data import, export, version management and historical operation record saving; Voronoi diagram generation module: through the built-in Voronoi diagram algorithm, based on the input cell center point coordinate data, the corresponding Voronoi space partition graph is generated; The Voronoi diagram generation module supports batch processing of multiple data sets; Cell score calculation module: based on the Voronoi polygon vertex data, the shape score of each cell is calculated; The calculation method combines the polygon area and the length ratio of the non-adjacent vertex diagonal, and quantifies the cell size and shape regularity; The cell score calculation module outputs cell score data for subsequent visualization and analysis; Data interaction and editing module: provide user interface, support dynamic editing of cell data, including deleting specified cells, undo operation and resetting data function; Support user input cell number for data filtering and operation; Visual display module: responsible for displaying Voronoi diagram and cell score through color mapping in user interface, supports interactive operation, provides multiple color mapping schemes and font size adjustment; Multiple data similarity comparison module: support importing multiple corneal endothelial cell data, based on cell spatial distribution, morphological characteristics, quantity ratio, density and spatial structure multidimensional index, calculate the similarity between data sets; Adopting weighted comprehensive algorithm, support user-defined weight parameter, realize flexible and accurate similarity evaluation; Provide numerical and graphical display of similarity results to assist clinical diagnosis and scientific research analysis.

2. A diagnostic method for predicting corneal endothelial function, characterized by, Comprise the following steps: Step A1: data acquisition and pretreatment; Collect the image data of corneal endothelial cells, use image processing algorithm to pretreat the original image, extract the spatial coordinate information of cell center point; Convert the extracted cell position data into point cloud data, store point set data file; Step A2: cell distribution analysis and Voronoi diagram construction; Through the import function, read the user selected JSON format corneal endothelial cell data file; Parse the JSON file, extract the cell number and corresponding two-dimensional coordinate data contained in it; Use the extracted coordinate data to call the built-in Voronoi diagram algorithm module to generate the corresponding Voronoi diagram; The generated Voronoi diagram is visualized in the core area of the software interface, and the cell area is distinguished by different colors, and the color mapping is based on cell score; Step A3: Voronoi polygon vertex data calculation cell score Based on the generated Voronoi polygon data, the morphological characteristics of each cell corresponding to the Voronoi region are quantified, and the cell score is calculated, which is used to reflect the size and shape regularity of the cell; The cell score calculation formula is as follows: Wherein, C is a constant, S is the polygon area, and R is the ratio of the maximum diagonal length to the minimum diagonal length of the non-adjacent vertex of the polygon; The polygon area S formula is as follows: wherein the ratio R of the maximum diagonal length D max and the minimum diagonal length D min of the non-adjacent vertices of the polygon is as follows: Step A4: interactive operation and data management; The interactive operation and data management specifically includes: Data deletion: Users can input specific cell numbers in the interface, and the system will parse the input and delete the corresponding cell data from the model. After the deletion operation, the system automatically updates the Voronoi diagram, recalculates the cell fraction, and displays the spatial distribution of the remaining cells in different colors or markers in the interface. Undo operation: The system automatically saves the current state, including cell positions, Voronoi polygons, fractions, and interface display parameters, into the operation history stack before each data change. When the user selects undo, the system pops out the last saved state from the operation history stack and restores to that state, ensuring that the user can safely undo mistakes or adjust analysis parameters. Reset operation: Users can choose to reset, and the system will clear all current analysis results and operation history, reload the original data, and restore to the initial state of the imported JSON file, returning to the initial model state. This operation is suitable for users who want to start the analysis process again. Data import: Users can import JSON-formatted corneal endothelial cell data files, and the system will parse the file content, extract cell numbers, positions, and morphological feature information, generate Voronoi diagrams and cell fractions, and update the interface display. Data export: After completing the current analysis, users can choose the format to save the spatial distribution, cell numbers, fractions, and Voronoi polygon information of the current model to a file, facilitating subsequent analysis or reporting. Step A5: Multi-group data similarity comparison Support importing multiple groups of JSON-formatted corneal endothelial cell spatial distribution data and using multiple indicators for multi-dimensional similarity evaluation.

3. The diagnostic method for predicting corneal endothelial function according to claim 2, characterized by: The multi-dimensional similarity evaluation of the multiple indicators in step A5 includes the following steps: Step Q1: KDE spatial distribution similarity, structural feature similarity, morphological feature similarity, graph theory feature similarity, and density similarity. The KDE spatial distribution similarity uses Gaussian kernel density estimation to model the spatial distribution of cell points, and then measures the overall spatial distribution consistency by calculating the similarity of the two distributions. The structural feature similarity analyzes the number of neighbors of each cell by constructing a Voronoi diagram, reflecting the local structural features of the organization. Morphological features: Extract the geometric features of each cell polygon to measure the similarity of cell shape. Graph theory feature similarity: Model the cell space as a graph and extract the structural features of the graph to measure the overall connection relationship of the organization. Density similarity: Compare the cell density in the effective area of the two samples to reflect the uniformity of cell distribution. Step Q2: Comprehensive similarity, weight the similarity of the above five indicators by pre-set or adjusted weights, and obtain the overall similarity. The specific similarity calculation formula is: Similarity Total = ω KDE × Similarity KDE + ω Structure × Similarity Structure + ω Morphology × Similarity Morphology + ω Graph × Similarity Graphy + ω Density × Similarity Density .

4. The diagnostic method for predicting corneal endothelial function according to claim 2, characterized by: The KDE spatial distribution similarity calculation step is as follows: Step S1: Use the spatial points of each group of cells as samples and use kernel density estimation to calculate the density values on the defined grid. where x represents a certain position point in the space, x i represents the ith sample point, n represents the total number of sample points, h represents a bandwidth parameter for controlling the smoothing degree of the kernel function, and K represents the kernel function. Step S2: Calculate the density distributions p(x, y) and q(x, y) of the two samples. Step S3: Use JS divergence to measure the difference between the two density distributions: wherein, D KL is the Kullback-Leibler divergence; Step S4: Similarity is: Similarity KDE = 1 - JSD(p, q); Structure feature similarity: By constructing a Voronoi diagram, the number of neighbors of each cell, i.e., the number of polygon vertices, is analyzed to reflect the local structure characteristics of the tissue.

5. The diagnostic method for predicting corneal endothelial function according to claim 2, characterized by: The structure feature similarity calculation step specifically is: Step D1, using a Voronoi diagram, the number of neighbors of each cell, i.e., the number of polygon vertices, is counted; Step D2, a histogram of the number of neighbors is constructed, the maximum number of neighbors is set to 10, and normalized to a probability distribution vector h1 and h2; Step D3, the cosine similarity of the two histograms is calculated:

6. The diagnostic method for predicting corneal endothelial function according to claim 2, characterized by: The morphological feature calculation step specifically is: Step F1, the area, perimeter, circularity, aspect ratio and Hu moment of each polygon are calculated; Step F2, each feature is normalized; Step F3, the Wasserstein distance WassersteinDistance(f1,f2) is used to measure the difference between the two feature distributions; Step F4, converted to similarity:

7. The diagnostic method for predicting corneal endothelial function according to claim 2, characterized by: The graph theory feature similarity calculation step specifically is: Step G1, construct a graph: nodes are cells, edges are adjacency relations, i.e., Delaunay triangulation; Step G2, extract features: degree distribution, clustering coefficient distribution, shortest path length distribution and global statistics; Step G3, calculate the L1 distance of each distribution: where p i and q i represent the component values of the two vectors in the i-th dimension, and n is the dimension of the vectors. Step G4, combined with the global statistics, the overall graph structure similarity is formed: where D i represents the difference measure of the i-th graph structure feature, ω i represents the weight coefficient of the corresponding feature.

8. The diagnostic method for predicting corneal endothelial function according to claim 2, characterized by: The density similarity calculation step specifically is: Step H1, count the number of cell points in the effective area of each group of samples; Step H2, calculate the density: Where Num represents the number of cells in the effective area, and S represents the area of the effective area; Step H3, compare the ratio of the two densities:

Images