Disease diagnosis method and system based on neural network cognitive diagnosis

Abstract

The invention relates to the technical field of disease diagnosis, and comprises a disease diagnosis method and system based on neural network cognitive diagnosis, and the method comprises the following steps: obtaining medical image data, calculating the gray level change rate, gradient direction distribution and edge continuity of a lesion region, counting the lesion tissue damage area, and analyzing the pathological tissue damage degree. And obtaining a lesion distribution consistency index. According to the invention, through comprehensive calculation of the medical image data and the pathological tissue slice data, fine-grained description of a focus area is realized, association among different lesion features is realized, and calculation of a lesion propagation path matching rate is realized, so that identification of lesion diffusion conditions is more accurate, and development trends of diseases in different tissue structures can be effectively predicted; the calculation of the cross-modal feature error is combined with the adjustment of the distribution weight of the lesion region, misdiagnosis caused by modal difference and extraction of abnormal signals in a lesion diffusion range are reduced, so that screening of potential diseases is more targeted, and the accuracy of disease recognition is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of disease diagnosis, and in particular, to a disease diagnosis method and system based on neural network cognitive diagnosis. Background Art

[0002] The technical field of disease diagnosis involves using various biomedical signals, imaging data, etc. to analyze the health status of patients to determine whether there are diseases or health abnormalities. The core content involves aspects such as data acquisition, feature extraction, pattern recognition, and clinical decision support. Data acquisition usually includes methods such as medical imaging scans, blood tests, and biosensor recordings to obtain the physiological or pathological information of patients. Feature extraction mainly analyzes the collected data to extract important indicators or biomarkers related to diseases. Pattern recognition usually relies on statistical analysis and machine learning methods to classify or predict data to assist doctors in making clinical judgments. Clinical decision support provides disease diagnosis suggestions or risk assessments by integrating various information resources to improve the accuracy of medical decisions.

[0003] Among them, the disease diagnosis method based on neural network cognitive diagnosis refers to using neural network technology to perform cognitive modeling on the medical data of patients to achieve automatic analysis and classification of diseases. Usually, a neural network model is trained based on a large-scale medical dataset so that it can identify the characteristic patterns of different diseases. Specifically, this method constructs a neural network structure, inputs information such as medical images, laboratory test data, or gene sequences, and uses hierarchical calculations to extract features and classify the data. During the training process, supervised learning or unsupervised learning strategies are adopted to enable the model to identify the potential features of different types of diseases. During the diagnosis process, after inputting the relevant data of the patient, the neural network outputs possible disease categories or diagnosis results through forward calculation and feature comparison. In addition, this method can also combine the patient's medical history information and adopt sequence analysis methods to improve the ability to understand the development trend of diseases.

[0004] The existing technology mainly relies on fixed rules for analysis in terms of lesion feature characterization, lacking in-depth calculations for multi-dimensional features of the lesion area, resulting in insufficient accuracy in identifying complex lesions. The prediction of the lesion diffusion trend relies on single-dimensional change information, making it difficult to establish a complete lesion propagation model and affecting the accurate judgment of the disease progression. Due to the failure to effectively match key parameters such as the lesion distribution pattern, abnormal signal range, and tissue damage gradient, the existing methods are prone to misdiagnosis due to insufficient feature matching during the disease classification process. The calculation of the matching error is usually based on a single data source, lacking unified calibration for multi-modal data, resulting in an error accumulation problem when matching image data and pathological data. The screening method for abnormal signals within the lesion diffusion range is relatively single, making it difficult to accurately extract the feature points of different diseases, affecting the adaptability of disease classification, and reducing the generalization ability for complex diseases. Summary of the Invention

[0005] The object of the present invention is to solve the drawbacks existing in the prior art, and a disease diagnosis method and system based on neural network cognitive diagnosis are proposed.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: A disease diagnosis method based on neural network cognitive diagnosis, comprising the following steps:

[0007] S1: Obtain medical image data, calculate the gray change rate, gradient direction distribution, and edge continuity of the lesion area, count the damaged area of the lesion tissue, analyze the degree of pathological tissue damage, and obtain the lesion distribution consistency index;

[0008] S2: Based on the lesion distribution consistency index, calculate the lesion influence degree between different regions, determine the lesion propagation priority, calculate the lesion diffusion change rate of the current patient, calculate the lesion propagation path matching rate, and obtain the lesion propagation mode matching result;

[0009] S3: Based on the lesion propagation mode matching result, calculate the matching degree of the disease type characteristics, extract the typical characteristics in the existing disease database, calculate the similarity with the current disease characteristics, and obtain the preliminary disease matching result;

[0010] S4: According to the preliminary disease matching result, calculate the cross-modal feature error of the matching area. If it exceeds the set range, adjust the lesion area distribution weight, calculate the area matching error, and obtain the lesion matching error correction record;

[0011] S5: Based on the lesion matching error correction record, analyze the diffusion trend of the current lesion, calculate the influence degree of the lesion in different tissue types, re-compare the existing disease database, and screen to obtain the disease recognition result.

[0012] As a further solution of the present invention, the lesion distribution consistency index includes gray change rate, gradient direction distribution, edge continuity, damaged area of the lesion tissue, and tissue destruction index; the lesion propagation mode matching result includes lesion influence degree, lesion propagation priority, lesion diffusion change rate, and lesion propagation path matching rate; the preliminary disease matching result includes disease type characteristic matching degree, lesion distribution pattern, abnormal signal range, tissue damage gradient, and preliminary disease classification result; the lesion matching error correction record includes cross-modal feature error, lesion area distribution weight, and area matching error; the disease recognition result includes lesion diffusion trend analysis result, lesion influence degree record, lesion diffusion range, abnormal signal record, and final disease determination result.

[0013] As a further solution of the present invention, the specific steps for obtaining medical image data, calculating the gray-scale change rate, gradient direction distribution, and edge continuity of the lesion area, counting the tissue damage area of the lesion, analyzing the degree of pathological tissue damage, and obtaining the lesion distribution consistency index are as follows:

[0014] S111: Obtain medical image data, including CT, MRI, and ultrasound images. For the lesion area in the image data, calculate the gray-scale value and gradient direction distribution of the lesion area, extract the coordinates of the edge pixel points, calculate the distance change between adjacent edge points, obtain the edge continuity parameter, call the gray-scale value, gradient direction distribution, and edge continuity parameter, compare the characteristic differences of different lesion areas, and generate the lesion area characteristic parameter;

[0015] S112: According to the lesion area characteristic parameter, obtain the pathological tissue section data, identify the tissue damage area of the lesion, calculate the tissue damage area within the lesion area, count the degree of damage of the regional pixel points, calculate the tissue damage index based on the tissue damage area and the degree of damage, call the gray-scale value of the lesion area, calculate the gray-scale change rate, analyze the numerical difference between the gray-scale change rate of the lesion area and the tissue damage index, and generate the lesion area damage parameter;

[0016] S113: According to the lesion area damage parameter, based on the difference between the gray-scale change rate of the lesion area and the tissue damage index, construct a feature map, obtain the mapping coordinates of different lesion areas in the feature space, calculate the consistency of the lesion distribution based on the mapping coordinates, count the matching degree of the lesion area in different image data and pathological sections, and calculate the lesion distribution consistency index.

[0017] As a further solution of the present invention, the specific steps for calculating the lesion influence degree between different regions, determining the lesion propagation priority, calculating the lesion diffusion change rate of the current patient, calculating the lesion propagation path matching rate, and obtaining the lesion propagation mode matching result based on the lesion distribution consistency index are as follows:

[0018] S211: Based on the lesion distribution consistency index, calculate the lesion influence degree between lesion areas, extract the spatial positions and characteristic parameters of each lesion area, calculate the gray-scale gradient difference between adjacent lesion areas, determine the lesion propagation priority, and obtain the lesion propagation priority parameter;

[0019] S212: According to the lesion propagation priority parameter, obtain the time series data of the lesion area of the current patient, calculate the change amplitude of the lesion area at consecutive time points, count the increase and decrease rate of the lesion area, calculate the lesion diffusion change rate based on the change trend of the lesion area, and obtain the lesion diffusion rate data;

[0020] S213: Obtain the existing disease transmission pattern data according to the lesion spread rate data, calculate the spatial matching degree between the lesion transmission trajectory and the existing disease transmission pattern, calculate the matching rate of the lesion transmission path based on the matching degree, and determine whether the matching rate of the lesion transmission path exceeds the set matching range. If the matching rate exceeds the range, associate the corresponding disease and obtain the matching result of the lesion transmission pattern.

[0021] As a further solution of the present invention, the matching rate is calculated using the formula:

[0022]

[0023] to calculate and determine whether the matching rate of the lesion transmission path exceeds the set matching range. If the matching rate exceeds the range, associate the corresponding disease. Among them, R p represents the matching rate of the lesion transmission path, d i represents the distance of lesion transmission in the i-th spatial unit, w i represents the weight factor of the i-th spatial unit, K represents the total number of spatial units in the lesion transmission path, s j represents the distance of the existing disease transmission pattern in the j-th spatial unit, v j represents the matching weight factor of the j-th spatial unit, and k represents the total number of spatial units in the existing disease transmission pattern.

[0024] As a further solution of the present invention, based on the matching result of the lesion transmission pattern, the specific steps to calculate the matching degree of disease type characteristics, extract the typical characteristics from the existing disease database, calculate the similarity with the current disease characteristics, and obtain the preliminary matching result of the disease are as follows:

[0025] S311: Based on the matching result of the lesion transmission pattern, extract the spatial topological structure of the lesion area, analyze the relative distribution relationship between the lesion areas, count the change in the lesion signal intensity of the differential lesion areas, determine the matching degree of the differential disease characteristics, and obtain the disease characteristic matching parameter data;

[0026] S312: According to the disease characteristic matching parameter data, extract the typical characteristics from the existing disease database, analyze the lesion spread trend based on the tissue damage gradient, and calculate the similarity of the current disease characteristics according to the lesion distribution pattern, abnormal signal range, and tissue damage gradient to obtain the disease characteristic similarity data;

[0027] S313: According to the disease characteristic similarity, calculate the potential disease classification based on the characteristic matching degree, screen the disease categories, count the relative matching rate of the differential disease classifications, and preliminarily determine the disease type according to the disease category with the highest matching rate to obtain the preliminary matching result of the disease.

[0028] As a further solution of the present invention, the similarity is calculated using the formula:

[0029]

[0030] Perform calculations, where S 病症 represents the disease feature similarity data, q i represents the weight of the i-th dimension of the disease feature, F i represents the value of the i-th dimension of the current disease feature data, X i represents the value of the i-th dimension feature of the matching disease in the reference disease database, max(F i , X i ) represents the maximum feature value of the current disease and the matching disease in the i-th dimension, Y i represents the distribution value of the disease feature in the i-th dimension, represents the average distribution value of all disease feature dimensions, and Q represents the total number of dimensions of the disease feature.

[0031] As a further solution of the present invention, according to the preliminary disease matching result, calculate the cross-modal feature error of the matching area. If it exceeds the set range, adjust the distribution weight of the lesion area, calculate the area matching error, and the specific steps for obtaining the lesion matching error correction record are as follows:

[0032] S411: Based on the preliminary disease matching result, extract the cross-modal feature data of the matching area, analyze the feature differences of the same lesion area in different modal images, calculate the cross-modal feature error. If the cross-modal feature error exceeds the set range, record the spatial coordinates of the error overrun area and obtain the cross-modal feature error value;

[0033] S412: According to the cross-modal feature error value, determine whether the error exceeds the set threshold. If the error is overrun, adjust the distribution weight of the lesion area, adjust the influence proportion of the lesion area in the disease matching calculation, re-distribute the weight ratio of the lesion area in the matching calculation, and obtain the lesion area distribution weight;

[0034] S413: Based on the lesion area distribution weight, extract the adjusted lesion area feature data, compare it with the matching data in the disease database, calculate the change of the adjusted matching error, count the distribution of the matching error of the different lesion areas, and generate a lesion matching error correction record.

[0035] As a further solution of the present invention, based on the lesion matching error correction record, analyze the diffusion trend of the current lesion, calculate the influence degree of the lesion in different tissue types, re-compare the existing disease database, and the specific steps for screening the disease recognition result are as follows:

[0036] S511: Based on the lesion matching error correction record, extract the time series data of the lesion area, count the spatial distribution of the lesion area at different time points, analyze the change range of the lesion boundary, count the change of signal intensity within the lesion area, and obtain the analysis result of the lesion diffusion trend;

[0037] S512: Based on the analysis result of the lesion diffusion trend, calculate the expansion ratio of the lesion signal in the normal tissue area, and count the influence degree of the lesion area in different tissue types according to the lesion diffusion rate of different tissue types, so as to obtain the lesion tissue influence record;

[0038] S513: According to the lesion tissue influence record, extract the abnormal signals within the lesion diffusion range, count the lesion signal distribution patterns of known diseases in the database, match them with the abnormal signals in the current lesion area, screen out the disease types that meet the characteristics, and screen out the disease categories that meet the standards according to the matching degree threshold, so as to obtain the disease recognition result.

[0039] A disease diagnosis system based on neural network cognitive diagnosis includes:

[0040] The lesion feature extraction module obtains the lesion area in the medical image data, extracts the gray change rate, gradient direction distribution, and edge continuity of the lesion area, calculates the damaged area of the lesion tissue in the pathological tissue section data, obtains the tissue damage index, calculates the difference between the gray change rate of the lesion area and the tissue damage index, establishes a feature mapping, and calculates the lesion distribution consistency index;

[0041] The lesion propagation evaluation module calculates the lesion influence degree based on the lesion distribution consistency index, determines the lesion propagation priority, calculates the lesion diffusion change rate, compares with the disease propagation mode data, calculates the lesion propagation path matching rate, and obtains the lesion propagation mode matching result;

[0042] The disease classification matching module calculates the matching degree of the disease type features based on the lesion propagation mode matching result, extracts the lesion distribution pattern, abnormal signal range, and tissue damage gradient, calculates the feature similarity, determines the potential disease classification, and obtains the preliminary disease matching result;

[0043] The lesion matching error correction module calculates the cross-modal feature error of the matching area based on the preliminary disease matching result, adjusts the distribution weight of the lesion area, and obtains the lesion matching error correction record;

[0044] The disease recognition module analyzes the lesion diffusion trend based on the lesion matching error correction record, extracts the abnormal signals within the lesion diffusion range, compares with the disease database and conducts screening to obtain the disease recognition result.

[0045] Compared with the prior art, the advantages and positive effects of the present invention are:

[0046] In the present invention, through the comprehensive calculation of medical image data and pathological tissue section data, a fine-grained description of the lesion area is achieved, making the quantitative analysis of the lesion area more complete, realizing the association between different lesion characteristics, enhancing the expression ability of complex pathological characteristics, calculating the matching rate of the lesion propagation path, making the identification of the lesion diffusion situation more accurate, and combining the determination of the lesion propagation priority, the development trend of the disease in different tissue structures can be effectively predicted. The feature extraction methods such as the lesion distribution pattern, abnormal signal range, and tissue damage gradient in the database enhance the comprehensiveness of the lesion feature matching and improve the adaptability to different disease types. In the matching error correction process, the calculation of the cross-modal feature error combined with the adjustment of the lesion area distribution weight makes the feature mapping between different image data and pathological data more accurate, reducing misdiagnosis caused by modal differences. The extraction of abnormal signals within the lesion diffusion range makes the screening of potential diseases more targeted. Combining the feature screening method of the database improves the accuracy of disease identification and enhances the adaptability to the individual pathological characteristics of different patients. Brief Description of the Drawings

[0047] Figure 1 It is a schematic diagram of the step flow of the present invention;

[0048] Figure 2 It is a flowchart of the steps of S1 of the present invention;

[0049] Figure 3 It is a flowchart of the steps of S2 of the present invention;

[0050] Figure 4 It is a flowchart of the steps of S3 of the present invention;

[0051] Figure 5 It is a flowchart of the steps of S4 of the present invention;

[0052] Figure 6 It is a flowchart of the steps of S5 of the present invention. Detailed Embodiment

[0053] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0054] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention. In addition, in the description of the present invention, the meaning of "a plurality" is two or more, unless otherwise specifically defined.

[0055] Please refer to Figure 1 , a disease diagnosis method based on neural network cognitive diagnosis, comprising the following steps:

[0056] S1: Obtain the lesion area in medical image data (CT, MRI, ultrasound), calculate the gray-scale change rate, gradient direction distribution, and edge continuity of the lesion area, obtain the damaged area of the lesion tissue in the pathological tissue section data, calculate the tissue damage index, analyze the degree of pathological tissue damage, and establish a feature map based on the difference between the gray-scale change rate of the lesion area and the tissue damage index to obtain the lesion distribution consistency index;

[0057] S2: Based on the lesion distribution consistency index, calculate the lesion influence degree between different regions, determine the priority of lesion propagation, calculate the change rate of lesion diffusion of the current patient, compare it with the existing disease propagation mode data, calculate the lesion propagation path matching rate, and if the matching rate exceeds the set range, associate the diseases to obtain the lesion propagation mode matching result;

[0058] S3: Based on the lesion propagation mode matching result, calculate the matching degree of disease type features, extract the typical features in the existing disease database, including the lesion distribution mode, abnormal signal range, and tissue damage gradient, calculate the feature similarity with the current disease, determine the potential disease classification, and obtain the preliminary disease matching result;

[0059] S4: According to the preliminary disease matching result, calculate the cross-modal feature error of the matching area. If it exceeds the set range, adjust the distribution weight of the lesion area, calculate the area matching error, and obtain the lesion matching error correction record;

[0060] S5: Based on the lesion matching error correction record, analyze the diffusion trend of the current lesion, calculate the influence degree of the lesion in different tissue types, extract the abnormal signals within the lesion diffusion range, re-compare the existing disease database, screen the diseases that meet the features, and obtain the disease recognition result.

[0061] The lesion distribution consistency indicators include the gray-scale change rate, gradient direction distribution, edge continuity, lesion tissue damage area, and tissue destruction index; the lesion propagation pattern matching results include the lesion influence degree, lesion propagation priority, lesion diffusion change rate, and lesion propagation path matching rate; the initial disease matching results include the matching degree of disease type characteristics, lesion distribution pattern, abnormal signal range, tissue damage gradient, and initial disease classification results; the lesion matching error correction record includes cross-modal feature errors, lesion area distribution weights, and regional matching errors; the disease recognition results include the lesion diffusion trend analysis results, lesion influence degree record, lesion diffusion range, abnormal signal record, and final disease determination result.

[0062] Please refer to Figure 2 , and the specific steps of S1 are as follows:

[0063] S111: Obtain medical image data, including CT, MRI, and ultrasound images. For the lesion areas in the image data, calculate the gray-scale values and gradient direction distributions of the lesion areas, extract the coordinates of the edge pixel points, calculate the distance changes between adjacent edge points, obtain the edge continuity parameters, and call the gray-scale values, gradient direction distributions, and edge continuity parameters to compare the characteristic differences of different lesion areas and generate lesion area characteristic parameters;

[0064] When obtaining medical image data, the original data sets of CT, MRI, and ultrasound images need to be called, and the lesion areas in the image data are partitioned to determine the coordinate ranges of each area. Based on the pixel matrix of the image, the pixel values of the lesion areas are extracted, and the gray-scale value distribution of the area is calculated. The gray-scale value range of the pixel points is set between 0 and 255, and the average gray-scale value can represent the basic brightness characteristics of the area. For example, if the gray-scale value range of a certain lesion area is 80 to 200, then its average gray-scale value is calculated as follows: Suppose the area contains n pixel points, and the gray-scale value of each pixel point is g i , then the average gray-scale value G is calculated as:

[0065]

[0066] If a certain lesion area contains 100 pixel points and the total sum of the pixel gray-scale values is 15000, then its average gray-scale value is 150. Further, obtain the gray-scale gradients of each pixel point. The calculation method is as follows: In the horizontal direction, calculate the gray-scale value difference between adjacent pixels. Let the gray-scale values of adjacent pixels be g x and g x+1 , then the gradient is calculated as follows:

[0067]

[0068] In the vertical direction, calculate the gray-scale gradients of adjacent pixels in the same way And calculate the gradient direction distribution based on the gradient direction. The gradient direction angle can be calculated by the following formula:

[0069]

[0070] If the horizontal gradient value of a certain pixel is 20 and the vertical gradient value is 10, then the gradient direction angle is approximately 26.6 degrees. Further extract the coordinates of the edge pixels, and determine the edge pixels by the way that the gray value change of adjacent pixels exceeds the set threshold. Set the edge gray value change threshold to 30. If the gray value change of adjacent pixels is greater than 30, then determine it as an edge pixel, record its coordinates, calculate the distance change between adjacent edge points, and calculate based on the Euclidean distance formula:

[0071]

[0072] If the coordinates of two edge points are (10, 10) and (13, 14) respectively, then the distance calculation is:

[0073]

[0074] Further count the edge continuity parameters of different lesion regions. Set the continuity threshold to 5. The determination of this threshold is based on the statistical results of the edge change characteristics of normal tissues, that is, within the non-lesion region, the adjacent distance of edge pixels usually fluctuates within the range of 3 to 6 pixels. Therefore, set 5 as the standard for edge continuity determination, and this value is adjusted according to the change of image resolution. For example, in a CT image with a higher resolution (512×512 pixels), this value can be appropriately increased to 6 to 8 pixels, while in an ultrasound image with a lower resolution (256×256 pixels), this value can be reduced to 4 pixels to adapt to the edge characteristics of different imaging methods. If the average distance between adjacent edge points in a certain lesion region is less than 5, then its edge is considered to be more continuous. Finally, call the gray value, gradient direction distribution and edge continuity parameters, compare the characteristic differences of different lesion regions, and obtain the lesion region characteristic parameters.

[0075] S112: According to the lesion region characteristic parameters, obtain the pathological tissue section data, identify the damaged area of the lesion tissue, calculate the tissue damage area within the lesion area, count the degree of damage of the regional pixel points, calculate the tissue damage index based on the tissue damage area and the degree of damage, call the gray value of the lesion region, calculate the gray change rate, analyze the numerical difference between the gray change rate of the lesion region and the tissue damage index, and generate the lesion region damage parameters;

[0076] Call the lesion region characteristic parameters, obtain the pathological tissue section data, and extract the damaged area of the lesion tissue in the section. Determine the pixel range of the damaged area according to the image segmentation method. Set the number of pixel points in the damaged area to n and the total pixel area to A. Then the damage area S is calculated as follows:

[0077]

[0078] If the total pixel area of a certain pathological section is 5000 and the number of pixel points in the damaged area is 1000, then the calculated damaged area is 0.2, that is, it accounts for 20% of the total tissue area. Further, the gray values of the pixel points in this area are statistically analyzed to calculate the damage degree of the tissue damaged area. It is set that the gray value lower than the set threshold (such as 60) is the severely damaged area. The number of pixel points m in the severely damaged area is counted, and the damage index BI is calculated:

[0079]

[0080] If a damaged area contains 1000 pixel points, and the severely damaged area accounts for 400 pixel points, then the damage index is 0.4. Based on the tissue damage area and the damage index, the comprehensive damage degree of the diseased tissue is calculated. The gray value of the diseased area is called, and the gray change rate is calculated. The initial gray value of a certain diseased area is set as G i , and the current gray value is G c , then the gray change rate R g is calculated as follows:

[0081]

[0082] If the initial gray value is 150 and the current gray value is 120, then the calculated gray change rate is:

[0083]

[0084] Further obtain the numerical difference between the gray change rate of the diseased area and the tissue damage index, and calculate its difference J:

[0085] J = |R g - BI|

[0086] If the gray change rate is 0.2 and the damage index is 0.4, then the calculated numerical difference is 0.2, and finally the damage parameters of the diseased area are obtained.

[0087] S113: According to the damage parameters of the diseased area, based on the difference between the gray change rate of the diseased area and the tissue damage index, construct a feature map, obtain the mapping coordinates of the differential diseased area in the feature space, calculate the consistency of the lesion distribution according to the mapping coordinates, statistically analyze the matching degree of the diseased area in different image data and pathological sections, and calculate the obtained lesion distribution consistency index;

[0088] Call the lesion area damage parameters, construct a feature map based on the difference between the gray change rate of the lesion area and the tissue destruction index, set the coordinate axes in the feature space, where the x-axis represents the gray change rate and the y-axis represents the tissue destruction index, map the features of the lesion area into this coordinate space. For example, if the gray change rate of a lesion area is 0.2 and the tissue destruction index is 0.4, then the mapped point coordinates are (0.2, 0.4). Obtain the mapped coordinates of all lesion areas in the feature space and calculate the Euclidean distance between different lesion areas. Set the feature coordinates of two lesion areas as (x1, y1) and (x2, y2) respectively, and calculate the distance D between the lesion areas:

[0089]

[0090] If the coordinates of two lesion areas are (0.2, 0.4) and (0.3, 0.5) respectively, then the distance calculation is as follows:

[0091]

[0092] Further calculate the consistency of the lesion distribution. Set the distribution consistency index CI as the average value of the distances between all pairs of lesion areas. The calculation formula is as follows:

[0093]

[0094] where N is the total number of lesion areas, D i is the feature distance of each lesion area. If a dataset contains 5 lesion areas and the total sum of the feature distances between them pairwise is 1.2, then the consistency index calculation is as follows:

[0095]

[0096] Finally, obtain the lesion distribution consistency index.

[0097] Please refer to Figure 3 , the specific steps of S2 are as follows:

[0098] S211: Based on the lesion distribution consistency index, calculate the lesion influence degree between lesion areas, extract the spatial positions and feature parameters of each lesion area, calculate the gray gradient difference between adjacent lesion areas, determine the lesion propagation priority, and obtain the lesion propagation priority parameter;

[0099] Based on the lesion distribution consistency index, obtain the spatial distribution information of the lesion areas, extract the central coordinates and areas of each lesion area, calculate the Euclidean distance between adjacent lesion areas, group the lesion areas according to the magnitude of the distance values to determine the lesion areas that influence each other, further extract the gray value distribution of each lesion area, and calculate the gray gradient change rate between adjacent lesion areas. For the areas where the gray gradient change rate is higher than the set gray gradient change threshold, calculate its gradient mean value and change amplitude, and judge the influence degree of the lesion according to the absolute value of the gradient change amplitude. For example, if the gray gradient mean value of a certain lesion area is 50 and the change amplitude is 20, while the gradient mean value of the adjacent lesion area is 45 and the change amplitude is 25, then calculate the influence degree of this area on the adjacent area as (50 - 45) + (20 - 25) = -5. If this value is less than the set influence threshold, the influence of this lesion area on the adjacent area is weak, otherwise it is strong. According to the influence degree values of each lesion area, compare the influence relationships between different areas, sort them according to the magnitude of the influence degree values, and determine the propagation priority of the lesions.

[0100] Basis for setting the influence threshold: Calculate based on the variation degree of the gray gradient of the lesion area and the image resolution. Specifically, it is set in the range of 20% to 30% of the gray gradient mean value of the image resolution. If the resolution of the image is 512×512 pixels and the gray value range is 0 - 255, then the gray gradient mean value can be calculated, and the set range is between (255 / 512)×20% and (255 / 512)×30%, that is, the influence threshold range is 0.1 to 0.15. The specific value is adjusted according to the average size of the lesion area. For example, if the average diameter of the lesion area is 50 pixels, the corresponding influence threshold is adjusted to 0.12, and if the average diameter of the lesion area is 100 pixels, the influence threshold is adjusted to 0.14, ensuring that the calculation of the influence degree between lesion areas conforms to their actual spatial scale. Finally, obtain the lesion propagation priority parameter.

[0101] S212: According to the lesion propagation priority parameter, obtain the time series data of the lesion areas of the current patient, calculate the change amplitude of the lesion areas at consecutive time points, count the increase and decrease rate of the lesion area area, and calculate the lesion diffusion change rate based on the change trend of the lesion area area to obtain the lesion diffusion rate data;

[0102] According to the lesion propagation priority parameter, obtain the imaging data of the current patient's lesion area at different time points, extract the coordinates of the lesion area in the time series, calculate the area of the lesion area at each time point, calculate the lesion diffusion change rate based on the increase or decrease of the lesion area, and calculate the ratio of the area difference between adjacent time points to the time interval. For example, if the area of the lesion area at time point t1 is 30 square millimeters, the area at time point t2 is 50 square millimeters, and the time interval is 5 days, then the calculated lesion diffusion change rate is (50 - 30) / 5 = 4 square millimeters per day. For the diffusion rates at different time points, calculate their average value and standard deviation to obtain the change trend of the lesion diffusion rate. Further extract the lesion propagation trajectory, connect the center coordinates of the lesion area at each time point to form the lesion diffusion path, and count the change of the diffusion direction of the lesion area along the path. For example, if the coordinates of the lesion area at three time points t1, t2, and t3 are (5, 5), (10, 10), and (15, 20) respectively, then the change rate of the lesion propagation direction is calculated as:

[0103] [(10 - 5) / (10 - 5) - (15 - 10) / (20 - 10)] / 2

[0104] Finally, obtain the lesion diffusion rate data.

[0105] S213: According to the lesion diffusion rate data, obtain the existing disease propagation mode data, calculate the spatial matching degree between the lesion propagation trajectory and the existing disease propagation mode, calculate the lesion propagation path matching rate based on the matching degree, and judge whether the lesion propagation path matching rate exceeds the set matching range. If the matching rate exceeds the range, then associate the corresponding disease to obtain the lesion propagation mode matching result;

[0106] The matching rate is calculated using the formula:

[0107]

[0108] Perform the calculation to judge whether the lesion propagation path matching rate exceeds the set matching range. If the matching rate exceeds the range, then associate the corresponding disease. Among them, R p represents the lesion propagation path matching rate, d i represents the distance of lesion propagation within the i-th spatial unit, w i represents the weight factor of the i-th spatial unit, K represents the total number of spatial units in the lesion propagation path, s j represents the distance of the existing disease propagation mode within the j-th spatial unit, v j represents the matching weight factor of the j-th spatial unit, and k represents the total number of spatial units in the existing disease propagation mode.

[0109] d i:The average gray value of the lesion area in the i-th spatial unit, which reflects the degree of lesion in this area. It is obtained by statistically calculating the pixel gray values of the lesion area in the medical image.

[0110] w i :The weight factor of the i-th spatial unit, indicating the importance of this area in the overall lesion propagation. It is determined according to factors such as the anatomical location and functional importance of this area. For example, areas of key organs may be given higher weights.

[0111] s j :The average gray value of the known disease propagation pattern in the j-th spatial unit, indicating the degree of lesion in this area in the known disease. It is obtained by analyzing and calculating historical medical image data.

[0112] v j :The matching weight factor of the j-th spatial unit, indicating the importance of this area in the known disease propagation pattern. It is determined according to the lesion frequency or severity of this area in the historical data.

[0113] K: The total number of spatial units in the new lesion propagation path.

[0114] k: The total number of spatial units in the known disease propagation pattern.

[0115] Average gray value d of the lesion area i and s j Calculation of:

[0116] Obtain image data: Obtain the image data of the patient through medical imaging equipment (such as MRI, CT).

[0117] Image segmentation: Use image segmentation algorithms (such as region growing, K-Means clustering, watershed algorithm, etc.) to divide the image into different spatial units and accurately locate the lesion area.

[0118] Gray value calculation: For the lesion area in each spatial unit, calculate the average of its pixel gray values to obtain d i and s j .

[0119] Determination of weight factors w i and v j :

[0120] Anatomical importance: According to medical knowledge, determine the anatomical importance of each spatial unit. For example, units involving important organs or functional areas should be given higher weights.

[0121] Lesion frequency: Analyze historical case data and count the frequency of lesions in each spatial unit. The higher the frequency, the larger the weight factor.

[0122] Calculate the weighted average gray value:

[0123] The weighted average gray value D of the new lesion avg :

[0124]

[0125] This value represents the weighted average gray value of the new lesion in each spatial unit and reflects the overall lesion degree.

[0126] The weighted average gray value S of the known disease avg :

[0127]

[0128] This value represents the weighted average gray value of the known disease in each spatial unit.

[0129] Calculate the lesion propagation path matching rate R p :

[0130] R p =|D avg -S avg |

[0131] This value represents the difference degree between the new lesion and the known disease in the propagation path. The smaller the value, the more similar the propagation patterns of the two are.

[0132] Suppose through medical image analysis, the following data are obtained:

[0133] New lesion data:

[0134] Spatial unit 1: average gray value d 1 =150, weight factor w 1 =1.5

[0135] Spatial unit 2: average gray value d 2 =200, weight factor w 2 =2.0

[0136] Spatial unit 3: average gray value d 3 =180, weight factor w 3 =1.8

[0137] Known disease data:

[0138] Spatial unit 1: average gray value s 1 =160, matching weight factor v 1 =1.4

[0139] Spatial unit 2: average gray value s 2 =190, matching weight factor v 2= 1.9

[0140] Spatial unit 3: average gray value s 3 = 175, matching weight factor v 3 = 1.7

[0141] Calculate the weighted average gray value D of the new lesion avg :

[0142]

[0143] Calculate the weighted average gray value S of the known disease avg :

[0144]

[0145] Calculate the lesion propagation path matching rate R p :

[0146] R p = |D avg - S avg |

[0147] = |179.06 - 176.5|

[0148] = 2.56

[0149] This result indicates that the average gray value of the new lesion differs from that of the known disease propagation pattern by 2.56. The difference in gray values reflects the matching situation of the lesion in spatial propagation. If the set matching threshold is 5, then this lesion propagation pattern is relatively similar to the known disease propagation pattern and may belong to the same or related diseases. If the difference exceeds the set threshold, then the new lesion propagation pattern may have different pathological characteristics and further analysis is required.

[0150] Please refer to Figure 4 , the specific steps of S3 are as follows:

[0151] S311: Based on the lesion propagation pattern matching result, extract the spatial topological structure of the lesion area, analyze the relative distribution relationship between the lesion areas, count the change in the lesion signal intensity of the differential lesion areas, determine the degree of feature matching of the differential diseases, and obtain the disease feature matching parameter data;

[0152] Based on the results of lesion propagation pattern matching, first obtain the distribution parameters of the lesion area, extract the spatial location, lesion size, and lesion density of the lesion area, count the relative distances of different lesion areas, and calculate the change in the degree of the lesion. If the degree of a certain lesion area has increased significantly compared to the surrounding areas, then this area may have a higher lesion influence ability. To accurately evaluate the lesion influence degree, it is necessary to calculate the gray gradient of the lesion area, extract the gray information of the lesion area, and analyze the change in the gray gradient of adjacent areas. If the gray gradient of a certain lesion area is higher than the set reference value, then it can be considered that the lesion activity in this area is high. At the same time, count the gray gradients of all lesion areas and calculate the average value as the reference data for lesion influence evaluation. According to the lesion influence degree, sort all lesion areas according to their influence ability. During the sorting process, give priority to marking the areas with higher lesion influence ability, and stratify and classify them according to the interaction between lesion areas to determine the lesion propagation priority. If the lesion influence ability of a certain area exceeds the average value of adjacent areas and its lesion trend shows a continuous upward state, then its leading role in lesion propagation can be further confirmed. After calculating the lesion propagation priority, form the priority arrangement of the lesion area, and finally obtain the disease feature matching parameter data.

[0153] S312: According to the disease feature matching parameter data, extract the typical features in the existing disease database, analyze the lesion diffusion trend based on the tissue damage gradient, and calculate the feature similarity of the current disease according to the lesion distribution pattern, abnormal signal range, and tissue damage gradient to obtain the disease feature similarity data;

[0154] Similarity, using the formula:

[0155]

[0156] Perform the calculation, where S 病症 represents the disease feature similarity data, q i represents the weight of the i-th dimension of the disease feature, F i represents the value of the i-th dimension of the current disease feature data, X i represents the value of the i-th dimension feature of the matching disease in the reference disease database, max(F i , X i ) represents the maximum feature value of the current disease and the matching disease in the i-th dimension, Y i represents the distribution value of the disease feature in the i-th dimension, represents the average distribution value of all disease feature dimensions, and Q represents the total number of dimensions of the disease feature.

[0157] Part 1: Weighted feature matching degree

[0158] This part measures the similarity between the current disease condition and the reference disease condition in each feature dimension, taking into account the weights of each feature.

[0159] q i : The weight of the i-th feature, which reflects the importance of this feature in the overall similarity calculation. The weights are set based on the impact of the features on disease diagnosis and are usually determined by expert experience or statistical analysis.

[0160] F i : The value of the current disease condition in the i-th feature dimension. This value is obtained by measuring or detecting the patient's clinical data.

[0161] X i : The value of the reference disease condition in the i-th feature dimension. This value usually comes from the average feature values of known disease conditions in the medical database.

[0162] Calculate the percentage difference between the current disease condition and the reference disease condition in each feature dimension:

[0163]

[0164] This formula calculates the relative difference between two feature values, and the result is between 0 and 1.

[0165] Calculate the similarity of each feature dimension:

[0166] Similarity = 1 - Percentage difference

[0167] The closer the similarity is to 1, the more similar the two feature values are.

[0168] Multiply the similarity by the weight of this feature:

[0169] Weighted similarity = w i × Similarity

[0170] Sum the weighted similarities of all feature dimensions and divide by the sum of the weights to obtain the weighted feature matching degree:

[0171]

[0172] Part Two: Standard Deviation of Feature Distribution

[0173] This part measures the dispersion degree of the current disease condition feature distribution, reflecting the fluctuation of feature values.

[0174] Y i : The value of the current disease condition in the i-th feature dimension, the same as F i Same.

[0175] The average value of all feature dimension values, calculated as:

[0176]

[0177] Q: The total number of feature dimensions.

[0178] Calculation process:

[0179] Calculate the squared difference between each eigenvalue and the average value: Sum the squared differences of all feature dimensions and divide by the total number of feature dimensions to obtain the variance: Take the square root of the variance to obtain the standard deviation: Add the weighted feature matching degree to the standard deviation of the feature distribution to obtain the final disease feature similarity S 病症 .

[0180] Suppose there are three feature dimensions (Q = 3), and their weights and values are as follows:

[0181] Feature 1: Weight q 1 = 0.5, current disease value F 1 = 80, reference disease value X 1 = 100;

[0182] Feature 2: Weight q 2 = 0.3, current disease value F 2 = 90, reference disease value X 2 = 85;

[0183] Feature 3: Weight q 3 = 0.2, current disease value F 3 = 70, reference disease value X 3 = 75;

[0184] Calculate the difference percentage:

[0185] Feature 1: Feature 2: Feature 3: Calculate the similarity:

[0186] Feature 1: 1 - 0.2 = 0.8

[0187] Feature 2: 1 - 0.0556 = 0.9444

[0188] Feature 3: 1 - 0.0667 = 0.9333

[0189] Calculate the weighted similarity:

[0190] Feature 1: 0.5 × 0.8 = 0.4

[0191] Feature 2: 0.3 × 0.9444 ≈ 0.2833

[0192] Feature 3: 0.2 × 0.9333 ≈ 0.1867

[0193] Calculate the weighted feature matching degree:

[0194]

[0195] Assume that the values Y of all feature dimensions i are equal to the feature values of the current disease:

[0196] Feature 1: Y 1 = 80

[0197] Feature 2: Y 2 = 90

[0198] Feature 3: Y 3 = 70

[0199] Calculate the feature mean:

[0200]

[0201] Calculate the square of the deviation of each feature value from the mean:

[0202] Feature 1: (80 - 80) 2 = 0

[0203] Feature 2: (90 - 80) 2 = 100

[0204] Feature 3: (70 - 80) 2 = 100

[0205] Calculate the variance:

[0206]

[0207] Calculate the standard deviation:

[0208]

[0209] Calculate the disease feature similarity:

[0210] S 病症 = 0.87 + 8.16 = 9.03

[0211] The calculated disease symptom feature similarity is 9.03, and this value is used to measure the degree of feature matching between the current disease and the reference disease. The higher the value, the higher the overall similarity of the features of the current disease to the reference disease. The weighted feature matching degree mainly reflects the similarity between different feature dimensions, while the standard deviation of the feature distribution is used to measure the degree of fluctuation of the disease symptoms features. By combining the calculations of these two, the similarity of the disease symptoms features can be evaluated more comprehensively, providing data support for disease matching and diagnosis.

[0212] S313: According to the disease symptom feature similarity, calculate the potential disease classification based on the degree of feature matching, screen the disease categories, count the relative matching rates of different disease classifications, and based on the disease category with the highest matching rate, preliminarily determine the disease type to obtain the preliminary disease matching result;

[0213] According to the disease symptom feature similarity, analyze the potential disease classifications of the current lesion area. Based on the matching degrees of different diseases, screen out the disease types with higher similarity. For this purpose, it is necessary to obtain the feature data of various diseases in the disease database and compare them with the feature data of the current lesion area, count the feature matching degrees of each disease. If the feature matching degree of a certain disease is higher than the set threshold, then this disease may be the potential classification of the current lesion area. To further confirm the accuracy of the matching disease, it is necessary to calculate the disease matching rate, count the matching probabilities of different diseases, and sort them according to the matching degree. If the matching rate of a certain disease is significantly higher than that of other diseases, then it can be determined that it is the most likely classification of the current lesion. After calculating the matching rates of all diseases, screen out the disease type with the highest matching degree and further confirm whether it meets the feature requirements of the lesion area, and finally determine the disease classification to obtain the preliminary disease matching result.

[0214] Please refer to Figure 5 , and the specific steps of S4 are as follows:

[0215] S411: Based on the preliminary disease matching result, extract the cross-modal feature data of the matching area, analyze the feature differences of the same lesion area in different modality images, calculate the cross-modal feature error. If the cross-modal feature error exceeds the set range, record the spatial coordinates of the error-exceeding area to obtain the cross-modal feature error value;

[0216] Based on the preliminary disease matching results, extract the cross-modal feature data of the matching region. According to the characteristic parameters of the lesion region in different modality images, extract the density information of CT images, the tissue contrast parameters of MRI images, and the echo intensity distribution of ultrasound images respectively. Statistically analyze the edge morphology and gray level gradient of the lesion region in each image modality, calculate the feature offset value of the lesion region in different image modalities. If the change in the gray level gradient of the same lesion region in different modality images exceeds the set range, this set range is determined based on the tissue type characteristics of the lesion region. For example, in soft tissue lesions, the change range of MRI signal contrast is usually between 5% and 15%. Considering the normal physiological fluctuations of tissues and the interference of image noise, the threshold range is set at

[0217] ±10%. That is, when the difference in gray level gradient of cross-modal image data exceeds 10%, it is considered that there is a significant error. Further analyze the spatial distribution of the error region, statistically analyze the spatial coordinates of the region where the error exceeds the limit, and calculate the proportion of the error region in the entire lesion region. This proportion threshold is determined based on the statistical analysis of the spatial distribution of the lesion region. For example, when the proportion of the lesion area is less than 10%, the local error has little impact on the overall matching. However, when the error distribution of the lesion region exceeds 20% of the overall lesion area, the cross-modal error may lead to a deviation in lesion matching. Therefore, the set proportion threshold is 20%. If the proportion of the error region exceeds this threshold, it is judged that the cross-modal feature error of this region has a greater impact, providing reference data for subsequent matching correction, and finally obtaining the cross-modal feature error value.

[0218] S412: According to the cross-modal feature error value, determine whether the error exceeds the set threshold. If the error exceeds the limit, adjust the distribution weight of the lesion region, adjust the influence proportion of the lesion region in the disease matching calculation, reallocate the weight ratio of the lesion region in the matching calculation, and obtain the distribution weight of the lesion region;

[0219] According to the cross-modal feature error value, judge whether the error exceeds the set threshold. If the error exceeds the limit, adjust the distribution weight of the lesion area. First, extract the lesion gray-scale information of the area where the error exceeds the limit, and count the gray-scale offset value of this area in different imaging modalities. If the gray-scale value of the lesion area is high in the CT image but shows a low signal in the MRI image, it may be a cross-modal error caused by tissue composition differences. Further analyze the structural boundary data of this area, count the boundary gradient value of the area where the error exceeds the limit, and compare it with the boundary gradient of the normal tissue area. The reference value is determined based on the boundary gradient statistical data of the tissue structure. For example, for soft tissue lesions with relatively clear lesion edges, their boundary gradients are usually between 0.1 and 0.2. To ensure that the boundary error of the lesion area does not overly affect the matching process, the reference value is set to 0.15. That is, when the boundary gradient deviation of the error area exceeds 0.15, it is determined that there are problems such as blurred lesion boundaries or morphological offsets in this area. If the gradient change exceeds this reference value, it is necessary to adjust the distribution weight of this area, recalculate the spatial weight coefficient of the lesion area, and reallocate the weight ratio of the lesion area according to the distribution of the cross-modal feature error. If the error in a certain area has a greater impact, reduce the weight of this area in the matching calculation, while increase the weight ratio of the area with a smaller error, so as to optimize the influence proportion of the lesion area in the disease matching calculation, and finally obtain the distribution weight of the lesion area.

[0220] S413: Based on the distribution weight of the lesion area, extract the adjusted lesion area feature data, compare it with the matching data in the disease database, calculate the change of the adjusted matching error, count the distribution of the matching errors of different lesion areas, and generate a lesion matching error correction record;

[0221] Calculate the matching error of the lesion area according to the distribution weight of the lesion area, extract the adjusted lesion area feature data, and compare it with the standard lesion features in the disease database to analyze the change of the lesion area matching error before and after adjustment. First, calculate the mean value of the matching error of the lesion area, and re-count the error of the area after error adjustment. If the matching error decreases, it means that the matching effect of the lesion area is optimized after the weight adjustment. If the error is still large, it is necessary to further optimize the weight distribution. At the same time, count the distribution of the matching errors of different lesion areas, analyze the impact of error correction on different areas, obtain the corrected error data, and finally establish a lesion matching error correction record.

[0222] Please refer to Figure 6 , the specific steps of S5 are as follows:

[0223] S511: Based on the lesion matching error correction record, extract the time series data of the lesion area, count the spatial distribution of the lesion area at different time points, analyze the change range of the lesion boundary, count the change of signal intensity within the lesion area, and obtain the analysis result of the lesion diffusion trend;

[0224] Based on the lesion matching error correction record, extract the time series data of the lesion area, obtain the spatial coordinates of the lesion area at different time points, and calculate the change range of the lesion boundary between adjacent time points to determine whether there is a continuous expansion or contraction trend of the lesion area in a specific direction. To this end, count the change of the lesion area and calculate the moving distance of the lesion center point on the time axis. Set the lesion center movement threshold according to the movement characteristics of the lesion type and anatomical region. If the lesion is located in a relatively stable soft tissue area, such as the liver or pancreas, its natural displacement within a short period of time usually does not exceed 1.5 mm. Therefore, set the movement of the lesion center exceeding 2 mm within a single time interval as abnormal diffusion behavior. This value changes with the lesion growth rate and tissue tension. If the surrounding tissue forms a large resistance to the lesion growth, this value may decrease, while in low-density tissues (such as lung tissue), this value may be slightly higher. Further analyze the morphological change of the lesion area boundary, extract the abnormal signal distribution in the lesion area, calculate the mean value and standard deviation of the abnormal signal intensity. If the intensity change of the abnormal signal exceeds 20%, and the signal area extends beyond the original lesion boundary, it indicates that the lesion signal diffusion trend is significant. Count the distribution density of the abnormal signal, calculate the diffusion influence range of the lesion according to the proportion of the abnormal signal in the lesion diffusion area, and evaluate the diffusion speed. If the diffusion speed exceeds the growth speed threshold of the stable lesion, for example, if the lesion area increases by more than 30% at three consecutive time points, it is determined that the lesion is in the accelerated diffusion stage. This reference value is derived from the lesion growth rate statistics. The actual measurement shows that the growth rate of most benign lesions is lower than 10% per time unit, while malignant lesions or inflammatory diffusion may exceed this reference. Finally, obtain the analysis result of the lesion diffusion trend.

[0225] S512: Based on the analysis result of the lesion diffusion trend, calculate the expansion ratio of the lesion signal in the normal tissue area, count the influence degree of the lesion area in different tissue types according to the lesion diffusion rate of different tissue types, and obtain the lesion tissue influence record;

[0226] Call the results of the lesion spread trend analysis, obtain the spatial information of the differential tissue type regions, extract the signal change data of the lesion regions in different tissue types, and calculate the spread ratio of the lesion signals in different tissues. First, count the distribution of the lesion regions in soft tissues, muscle tissues, and adipose tissues, and extract the lesion signal characteristics of each tissue type. Calculate the attenuation or enhancement amplitude of the lesion signals in different tissue environments. For example, in adipose tissue, the lesion signal may attenuate by 20%, while around vascular tissue, the signal may enhance by 15%. Set the benchmark value of the lesion region expansion rate in muscle tissue based on the physiological characteristics and lesion spread pattern of muscle tissue. In muscle tissue, inflammatory lesions usually have a faster spread rate, up to 5 mm / day, while the spread rate of low-invasive tumors is less than 2 mm / day. Therefore, if the lesion spread rate exceeds 5 mm / day, the lesion may have strong invasiveness. This value fluctuates with the blood supply, cell proliferation rate, and matrix environment of muscle tissue. Lesions spread faster in regions with rich blood supply and slower in regions with dense muscle bundles. If the spread rate of the lesion signal in a specific tissue type exceeds the set benchmark, for example, if the lesion region expansion rate in muscle tissue exceeds 5 mm / day, it indicates that this tissue type has a greater impact on lesion spread. Further count the abnormal signal boundaries of the lesion regions and calculate the spread ratio of the lesion signals in normal tissues. If the coverage area of the abnormal signal in the surrounding tissues exceeds 30% of the original lesion region, it is considered that the lesion signal has strong spread ability in this tissue. Based on the signal propagation of the lesion in different tissue types, calculate the spread influence degree of the lesion tissue and establish a tissue distribution model of the lesion signal to evaluate the growth pattern of the lesion in different tissue types, and finally obtain the lesion tissue impact record.

[0227] S513: According to the lesion tissue impact record, extract the abnormal signals within the lesion spread range, count the lesion signal distribution patterns of known diseases in the database, match them with the abnormal signals of the current lesion region, screen out the disease types that meet the characteristics, and screen out the disease categories that meet the standards according to the matching degree threshold to obtain the disease recognition result;

[0228] Call the lesion tissue impact record, extract the abnormal signals within the lesion diffusion range, obtain the lesion feature data in the disease database, screen the disease types in the database that are similar to the current lesion area features, and calculate the matching degree. First, extract the lesion signal distribution patterns of all known diseases in the database, obtain their diffusion rates and signal attenuation characteristics in different tissue types, and compare them with the diffusion pattern of the current lesion area to calculate the feature similarity of the lesion signals. Set the similarity threshold of the lesion signals based on the signal matching conditions of different diseases in the database. Under normal circumstances, the matching degree of the lesion signals is between 75% and 95%. Among them, the diseases with a similarity higher than 85% usually belong to the highly suspected category. Therefore, set the similarity of the lesion signals to be greater than 85% as the matching standard, and this value fluctuates with the changes in the lesion type, imaging modality, and signal extraction method. For example, in CT images, due to the obvious gray-scale difference, the matching threshold can be reduced to 80%, while in MRI images, due to the influence of soft tissue contrast, the matching degree requirement is usually higher, reaching more than 90%. If the matching degree of a certain disease exceeds the set threshold, for example, the similarity of the lesion signals is greater than 85%, then this disease may be the matching category of the current lesion. To further screen the disease types that meet the characteristics, calculate the signal matching weights of different diseases, sort them according to the signal matching degree, screen out the disease with the highest matching degree, and analyze its adaptability in the lesion area. If the signal distribution pattern of the highest matching disease has a high consistency with the abnormal signal distribution of the current lesion area, then finally determine this disease as the recognition result of the current lesion, and finally obtain the disease recognition result.

[0229] A disease diagnosis system based on neural network cognitive diagnosis, including:

[0230] The lesion feature extraction module obtains the lesion area in the medical image data, extracts the gray-scale change rate, gradient direction distribution, and edge continuity of the lesion area, calculates the lesion tissue damage area in the pathological tissue section data, obtains the tissue damage index, calculates the difference between the gray-scale change rate of the lesion area and the tissue damage index, establishes a feature mapping, and calculates the lesion distribution consistency index;

[0231] The lesion propagation evaluation module calculates the lesion impact degree based on the lesion distribution consistency index, determines the lesion propagation priority, calculates the lesion diffusion change rate, compares the disease propagation mode data, calculates the lesion propagation path matching rate, and obtains the lesion propagation mode matching result;

[0232] The disease classification matching module calculates the feature matching degree of the disease type based on the lesion propagation mode matching result, extracts the lesion distribution pattern, abnormal signal range, and tissue damage gradient, calculates the feature similarity, determines the potential disease classification, and obtains the preliminary disease matching result;

[0233] The lesion matching error correction module calculates the cross-modal feature error of the matching region based on the preliminary disease matching result, adjusts the distribution weight of the lesion region, and obtains the lesion matching error correction record;

[0234] The disease recognition module analyzes the lesion diffusion trend based on the lesion matching error correction record, extracts the abnormal signals within the lesion diffusion range, compares with the disease database and conducts screening to obtain the disease recognition result.

[0235] The above are only the preferred embodiments of the present invention, and do not limit the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A disease diagnosis method based on neural network cognitive diagnosis, characterized in that: The following steps are involved: S1: Obtain medical imaging data, calculate the grayscale change rate, gradient direction distribution, and edge continuity of the lesion area, count the lesion tissue damage area, analyze the degree of pathological tissue damage, and obtain the lesion distribution consistency index; S2: Based on the lesion distribution consistency index, calculate the lesion influence degree between the difference areas, determine the lesion propagation priority, calculate the current patient lesion diffusion change rate, calculate the lesion propagation path matching rate, and obtain the lesion propagation mode matching result; S3: Based on the lesion propagation pattern matching result, the degree of matching of the disease type characteristics is calculated, typical characteristics in the existing disease database are extracted, the similarity with the characteristics of the current disease is calculated, and the preliminary matching result of the disease is obtained; S4: Calculate the cross-modal feature error of the matching area according to the preliminary matching result of the disease. If it exceeds the set range, adjust the distribution weight of the lesion area, calculate the regional matching error, and obtain the lesion matching error correction record; S5: Based on the lesion matching error correction record, analyze the diffusion trend of the current lesion, calculate the impact of the lesion in the different tissue types, re-compare with the existing disease database, and screen to obtain the disease recognition result.

2. The disease diagnosis method based on neural network cognitive diagnosis according to claim 1, characterized in that: The lesion distribution consistency indicators include grayscale change rate, gradient direction distribution, edge continuity, lesion tissue damage area, and tissue destruction index; the lesion propagation pattern matching results include lesion impact degree, lesion propagation priority, lesion diffusion change rate, and lesion propagation path matching rate; the preliminary disease matching results include disease type feature matching degree, lesion distribution pattern, abnormal signal range, tissue damage gradient, and preliminary disease classification results; the lesion matching error correction record includes cross-modal feature error, lesion area distribution weight, and regional matching error; the disease identification results include lesion diffusion trend analysis results, lesion impact degree records, lesion diffusion range, abnormal signal records, and final disease determination results.

3. The disease diagnosis method based on neural network cognitive diagnosis according to claim 1, characterized in that: The specific steps for obtaining medical imaging data, calculating the grayscale change rate, gradient direction distribution, and edge continuity of the lesion area, counting the lesion tissue damage area, analyzing the degree of pathological tissue damage, and obtaining the lesion distribution consistency index are as follows: S111: Obtain medical imaging data, including CT, MRI and ultrasound images, calculate the grayscale value and gradient direction distribution of the lesion area in the image data, extract edge pixel coordinates, calculate the distance change between adjacent edge points, obtain edge continuity parameters, call the grayscale value, gradient direction distribution and edge continuity parameters, compare the feature differences of different lesion areas, and generate lesion area feature parameters; S112: acquiring pathological tissue slice data according to the characteristic parameters of the lesion area, identifying the lesion tissue damage area, calculating the tissue damage area in the lesion area, counting the damage degree of the regional pixels, calculating the tissue damage index according to the tissue damage area and damage degree, calling the gray value of the lesion area, calculating the gray change rate, analyzing the numerical difference between the gray change rate of the lesion area and the tissue damage index, and generating the lesion area damage parameter; S113: According to the damage parameters of the lesion area and the difference between the grayscale change rate of the lesion area and the tissue destruction index, a feature mapping is constructed to obtain the mapping coordinates of the differential lesion area in the feature space, the consistency of the lesion distribution is calculated based on the mapping coordinates, the matching degree of the lesion area in different image data and pathological sections is statistically analyzed, and the lesion distribution consistency index is calculated.

4. The disease diagnosis method based on neural network cognitive diagnosis according to claim 1, characterized in that: Based on the lesion distribution consistency index, the lesion influence degree between the difference areas is calculated, the lesion propagation priority is determined, the current patient lesion diffusion change rate is calculated, the lesion propagation path matching rate is calculated, and the specific steps of obtaining the lesion propagation pattern matching result are as follows: S211: Based on the lesion distribution consistency index, calculate the lesion influence degree between lesion areas, extract the spatial position and characteristic parameters of each lesion area, calculate the gray gradient difference between adjacent lesion areas, determine the lesion propagation priority, and obtain the lesion propagation priority parameter; S212: according to the lesion propagation priority parameter, obtaining the time series data of the lesion area of ​​the current patient, calculating the change amplitude of the lesion area at consecutive time points, counting the increase and decrease rate of the lesion area, calculating the lesion diffusion change rate according to the lesion area change trend, and obtaining the lesion diffusion rate data; S213: According to the lesion diffusion rate data, the existing disease propagation pattern data is obtained, the spatial matching degree between the lesion propagation trajectory and the existing disease propagation pattern is calculated, the lesion propagation path matching rate is calculated based on the matching degree, and it is determined whether the lesion propagation path matching rate exceeds the set matching range. If the matching rate exceeds the range, the corresponding disease is associated to obtain the lesion propagation pattern matching result.

5. The disease diagnosis method based on neural network cognitive diagnosis according to claim 4 is characterized in that: The matching rate is calculated using the formula: Calculate and determine whether the lesion propagation path matching rate exceeds the set matching range. If the matching rate exceeds the range, the corresponding disease is associated, where R p represents the lesion propagation path matching rate, d i represents the distance of lesion propagation in the i-th spatial unit, w i represents the weight factor of the ith spatial unit, K represents the total number of spatial units in the lesion propagation path, and s j represents the distance of the existing disease propagation pattern within the jth spatial unit, v j represents the matching weight factor of the jth spatial unit, and k represents the total number of spatial units in the existing disease transmission pattern.

6. The disease diagnosis method based on neural network cognitive diagnosis according to claim 1, characterized in that: Based on the lesion propagation pattern matching result, the specific steps of calculating the matching degree of the disease type characteristics, extracting the typical characteristics in the existing disease database, calculating the similarity with the characteristics of the current disease, and obtaining the preliminary matching result of the disease are as follows: S311: based on the lesion propagation pattern matching result, extract the spatial topological structure of the lesion area, analyze the relative distribution relationship between the lesion areas, count the changes in the lesion signal intensity of the different lesion areas, determine the feature matching degree of the different diseases, and obtain the disease feature matching parameter data; S312: extracting typical features from an existing disease database according to the disease feature matching parameter data, analyzing the lesion diffusion trend according to the tissue damage gradient, calculating the feature similarity of the current disease according to the lesion distribution pattern, the abnormal signal range and the tissue damage gradient, and obtaining disease feature similarity data; S313: According to the similarity of the disease characteristics, the potential disease classification is calculated according to the degree of feature matching, the disease category is screened, the relative matching rate of the difference disease classification is counted, and the disease type is preliminarily determined according to the disease category with the highest matching rate to obtain the preliminary disease matching result.

7. The disease diagnosis method based on neural network cognitive diagnosis according to claim 6, characterized in that: The similarity is calculated using the formula: Calculate, where S 病症 Represents the disease feature similarity data, q i represents the weight of the i-th dimension of the disease feature, F i Represents the value of the i-th dimension of the current disease feature data, X i represents the feature value of the i-th dimension of the matching disease in the reference disease database, max(F i ,X i ) represents the maximum eigenvalue of the current disease and the matching disease in the i-th dimension, Y i Represents the distribution value of the disease characteristics in the i-th dimension, represents the average distribution value of all disease feature dimensions, and Q represents the total number of disease feature dimensions.

8. The disease diagnosis method based on neural network cognitive diagnosis according to claim 1, characterized in that: According to the preliminary matching results of the disease, the cross-modal feature error of the matching area is calculated. If it exceeds the set range, the distribution weight of the lesion area is adjusted, the regional matching error is calculated, and the specific steps of obtaining the lesion matching error correction record are as follows: S411: based on the preliminary matching result of the disease, extracting cross-modal feature data of the matching area, analyzing the feature difference of the same lesion area in the difference modality images, calculating the cross-modal feature error, and if the cross-modal feature error exceeds the set range, recording the spatial coordinates of the error exceeding limit area, and obtaining the cross-modal feature error value; S412: judging whether the error exceeds a set threshold value according to the cross-modal feature error value, and if the error exceeds the threshold, adjusting the distribution weight of the lesion area, adjusting the influence proportion of the lesion area in the disease matching calculation, reallocating the weight proportion of the lesion area in the matching calculation, and obtaining the distribution weight of the lesion area; S413: Based on the lesion area distribution weight, extract the adjusted lesion area feature data, compare it with the matching data in the disease database, calculate the change of the adjusted matching error, count the distribution of the difference lesion area matching errors, and generate a lesion matching error correction record.

9. The disease diagnosis method based on neural network cognitive diagnosis according to claim 1, characterized in that: Based on the lesion matching error correction record, analyzing the diffusion trend of the current lesion, calculating the influence of the lesion in the different tissue types, re-comparing the existing disease database, and screening to obtain the disease recognition results are as follows: S511: based on the lesion matching error correction record, extracting time series data of the lesion area, counting the spatial distribution of the lesion area at different time points, analyzing the change amplitude of the lesion boundary, counting the change of signal intensity within the lesion area, and obtaining the lesion diffusion trend analysis result; S512: Based on the lesion diffusion trend analysis result, the expansion ratio of the lesion signal in the normal tissue area is calculated, and according to the lesion diffusion rate of the different tissue types, the influence degree of the lesion area in the different tissue types is counted to obtain the lesion tissue influence record; S513: According to the diseased tissue impact record, the abnormal signals within the lesion diffusion range are extracted, the lesion signal distribution patterns of known diseases in the database are statistically analyzed, and the abnormal signals of the current lesion area are matched, and the disease types that meet the characteristics are screened. According to the matching degree threshold, the disease categories that meet the standards are screened to obtain the disease recognition results.

10. A disease diagnosis system based on neural network cognitive diagnosis, characterized in that: According to any one of claims 1 to 9, the disease diagnosis method based on neural network cognitive diagnosis comprises: The lesion feature extraction module obtains the lesion area in the medical imaging data, extracts the grayscale change rate, gradient direction distribution, and edge continuity of the lesion area, calculates the lesion tissue damage area in the pathological tissue section data, obtains the tissue destruction index, calculates the difference between the grayscale change rate of the lesion area and the tissue destruction index, establishes feature mapping, and calculates the lesion distribution consistency index; The lesion propagation evaluation module calculates the lesion impact degree, determines the lesion propagation priority, calculates the lesion diffusion change rate, compares the disease propagation pattern data, calculates the lesion propagation path matching rate, and obtains the lesion propagation pattern matching result based on the lesion distribution consistency index; The disease classification matching module calculates the degree of disease type feature matching based on the lesion propagation pattern matching result, extracts the lesion distribution pattern, abnormal signal range, tissue damage gradient, calculates feature similarity, determines potential disease classification, and obtains preliminary disease matching results; The lesion matching error correction module calculates the cross-modal feature error of the matching area based on the preliminary matching result of the disease, adjusts the distribution weight of the lesion area, and obtains the lesion matching error correction record; The symptom identification module analyzes the lesion diffusion trend based on the lesion matching error correction record, extracts abnormal signals within the lesion diffusion range, compares and screens the symptom database, and obtains symptom identification results.

Images