An artificial intelligence-based auxiliary system for ovarian cancer risk assessment

Abstract

The present invention discloses an artificial intelligence-based ovarian cancer risk assessment auxiliary system, comprising a data acquisition module, an ovarian feature extraction module, an ovarian cancer risk assessment model construction module, a model performance improvement module, and an ovarian cancer risk assessment auxiliary module. The present invention relates to the field of cancer risk assessment technology, and specifically refers to an artificial intelligence-based ovarian cancer risk assessment auxiliary system. This solution improves the accuracy and comprehensiveness of feature extraction through detail enhancement, contour extraction, morphological feature calculation, and texture feature calculation; improves the classification performance and stability of the model, and enhances the accuracy and reliability of ovarian cancer risk assessment, by constructing a double-layer fusion kernel function, designing an innovative penalty term, generating an objective function, and determining optimal conditions; and improves the efficiency and stability of model performance improvement by constructing a diversity maintenance factor, designing a parameter competition mechanism, compensating for step lengths, and performing crossover mutation.

Description

Technical Field

[0001] The present invention relates to the technical field of cancer risk assessment, and in particular to an artificial intelligence-based auxiliary system for ovarian cancer risk assessment. Background Art

[0002] An AI-based ovarian cancer risk assessment assistance system utilizes advanced AI technology and historical patient data to accurately assess ovarian cancer risk. By building an AI model and comprehensively analyzing various patient data, the system predicts ovarian cancer risk, provides decision support to clinicians, and facilitates early detection and intervention of ovarian cancer, thereby improving patient survival and quality of life.

[0003] Traditional ovarian feature extraction methods have problems such as loss of details, inaccurate contour extraction, incomplete feature extraction and insufficient data utilization; traditional ovarian cancer risk assessment model construction methods have problems such as single kernel function, unreasonable penalty terms, incomplete objective function and unclear optimal conditions; traditional model performance improvement methods have problems such as unclear parameter performance evaluation, insufficient diversity maintenance, imperfect competition mechanism, unreasonable compensation step size and insufficient crossover mutation. Summary of the Invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides an artificial intelligence-based ovarian cancer risk assessment auxiliary system. In view of the problems of detail loss, inaccurate contour extraction, incomplete feature extraction and insufficient data utilization in traditional ovarian feature extraction methods, this solution improves the accuracy and comprehensiveness of feature extraction through detail enhancement, accurate contour extraction, comprehensive morphological feature calculation, texture feature calculation, feature integration and data integration, provides richer and more effective data support for model training, thereby improving the accuracy and reliability of ovarian cancer risk assessment; in view of the problems of single kernel function, unreasonable penalty term and unreasonable objective function in traditional ovarian cancer risk assessment model construction method, To address the problem of unclear comprehensive and optimal conditions, this solution improves the model construction quality, classification performance and stability by constructing a double-layer fusion kernel function, designing innovative penalty terms, generating objective functions and determining optimal conditions, thereby improving the accuracy and reliability of ovarian cancer risk assessment. To address the problems of unclear parameter performance evaluation, insufficient diversity maintenance, imperfect competition mechanism, unreasonable compensation step size and insufficient cross-mutation in traditional model performance improvement methods, this solution improves the efficiency and stability of model performance improvement by constructing diversity maintenance factors, designing parameter competition mechanisms, designing compensation step sizes and performing cross-mutation, thereby improving the accuracy and reliability of the ovarian cancer risk assessment model.

[0005] The technical solution adopted by the present invention is as follows: The present invention provides an artificial intelligence-based ovarian cancer risk assessment auxiliary system, including a data acquisition module, an ovarian feature extraction module, an ovarian cancer risk assessment model construction module, a model performance improvement module and an ovarian cancer risk assessment auxiliary module;

[0006] The data acquisition module collects clinical data, imaging data, laboratory test data and final disease data of historical patients;

[0007] The ovarian feature extraction module extracts ovarian features by enhancing details, extracting ovarian contours, extracting morphological features, extracting texture features, generating feature vectors, and constructing a model training data set;

[0008] The ovarian cancer risk assessment model construction module constructs the ovarian cancer risk assessment model by setting labels, constructing a double-layer fusion kernel function, designing penalty terms, generating an objective function, and determining optimal conditions;

[0009] The model performance improvement module improves model performance through initialization, parameter performance evaluation, construction of diversity maintenance factors, design of parameter competition mechanism, design of compensation step size, crossover mutation and determination process;

[0010] The ovarian cancer risk assessment auxiliary module utilizes an ovarian cancer risk assessment model to assist in assessing a patient's ovarian cancer risk.

[0011] Furthermore, the data acquisition module collects clinical data, imaging data, laboratory test data and final disease data of historical patients; the clinical data include the patient's age, family medical history, number of pregnancies, whether there is abdominal pain and whether there is abdominal distension; the imaging data refers to the patient's ultrasound examination image; the laboratory test data is the test results of tumor markers, blood routine, liver function, kidney function and hormone levels; the final disease data refers to whether the patient has ovarian cancer.

[0012] Furthermore, the ovarian feature extraction module specifically includes the following contents:

[0013] Detail enhancement: weighted superposition and Laplacian operator enhancement are performed on the original image. By adaptively adjusting parameters, linear grayscale transformation is combined with normalized detail highlighting.

[0014] Extract the ovarian contour by calculating the horizontal and vertical gradient magnitudes and directions on the enhanced image. Using edge strength and direction information, the edge pixels that are consistent with the direction of the ovarian contour are selectively retained to extract the ovarian contour.

[0015] Morphological feature extraction: calculating the area, perimeter, major axis, and minor axis of the contour area, and then calculating the compactness and aspect ratio to quantitatively describe the regularity and stretching degree of the ovarian shape;

[0016] Texture feature extraction: Calculate the image's contrast, correlation, energy, and homogeneity based on the gray-level co-occurrence matrix to characterize the grayscale distribution and spatial relationship of the ovarian region.

[0017] Feature vector generation, which combines the contrast, correlation, energy, homogeneity, compactness and aspect ratio of the ovarian contour image into an ovarian feature vector;

[0018] Construct a model training data set, add the patient's clinical data, laboratory test data and final disease data to the ovarian feature vector to form a final data vector, create a model training data set, and add the final data vector to the model training data set.

[0019] Furthermore, the construction of the ovarian cancer risk assessment model module specifically includes the following contents:

[0020] Set labels and set the final diseased data as label data for the ovarian cancer risk assessment model;

[0021] Construct a double-layer fusion kernel function by fusing the Gaussian kernel function and the normalization part based on the inverse matrix of the covariance matrix;

[0022] Design the penalty term, consider the spatial smoothness of the kernel function, introduce the weighted sum of the squared gradient norm, and constrain the complexity of the model;

[0023] Generate the objective function. On the basis of maximizing the Lagrangian dual objective, add the quadratic penalty term and the kernel smoothing penalty to construct the objective function.

[0024] Determine the optimal conditions, and through the Lagrangian dual conditions, make the gradient of the Lagrangian function with respect to weights and biases zero, and at the same time satisfy the non-negativity of the Lagrangian multiplier, the classification interval constraint and the complementary relaxation.

[0025] Furthermore, the model performance improvement module specifically includes the following contents:

[0026] Initialization, determine the search parameters, including bandwidth parameters and scaling factors, and randomly generate initial parameter search points;

[0027] Parameter performance evaluation: set the model accuracy as the parameter performance value and set the parameter performance target value;

[0028] Construct a diversity maintenance factor, which is obtained by quantifying the breadth of parameter distribution through the average value of the ratio of the range of the search point set in each dimension to the corresponding search width;

[0029] Design a parameter competition mechanism, generate a competition probability by scaling the average of the best and worst performance values with a diversity factor, and use this probability to weight the current, best, worst, and average positions to update the parameter point position;

[0030] Design the compensation step size, which is calculated based on the ratio of the search point performance standard deviation to the average performance, and the difference between the average position of the initial point and the position of the worst performance point;

[0031] Cross-mutation, using Gaussian perturbation and differential mutation strategies on parameter points, using compensation steps and randomly selected other search points to generate new candidate solutions;

[0032] Determine the process and set the process of using the parameter competition mechanism to update the position of the parameter point and then perform a cross mutation as a search process; each time a search is performed, calculate the parameter performance value of the parameter search point after the search. If the parameter performance value is greater than the parameter performance standard value, the search stops and the parameter is qualified; otherwise, the parameter is unqualified and the next search is performed.

[0033] Furthermore, the ovarian cancer risk assessment auxiliary module collects the user's clinical data, imaging data and laboratory test data, first extracts ovarian features, and then inputs the extracted data into the ovarian cancer risk assessment auxiliary model. The model predicts the user's final disease status. If the prediction result is that the user is ill, the ovarian cancer risk is high risk. If the prediction result is that the user is not ill, the ovarian cancer risk is low risk.

[0034] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0035] (1) In response to the problems of detail loss, inaccurate contour extraction, incomplete feature extraction and insufficient data utilization in traditional ovarian feature extraction methods, this scheme improves the accuracy and comprehensiveness of feature extraction through detail enhancement, accurate contour extraction, comprehensive morphological feature calculation, texture feature calculation, feature integration and data integration, providing richer and more effective data support for model training, thereby improving the accuracy and reliability of ovarian cancer risk assessment.

[0036] (2) In response to the problems of single kernel function, unreasonable penalty term, incomplete objective function and unclear optimal conditions in the traditional ovarian cancer risk assessment model construction method, this scheme improves the model construction quality, classification performance and stability by constructing a double-layer fusion kernel function, designing innovative penalty terms, generating objective function and determining optimal conditions, thereby improving the accuracy and reliability of ovarian cancer risk assessment.

[0037] (3) In response to the problems of unclear parameter performance evaluation, insufficient diversity maintenance, imperfect competition mechanism, unreasonable compensation step size and insufficient cross-mutation in traditional model performance improvement methods, this scheme improves the efficiency and stability of model performance improvement by constructing diversity maintenance factors, designing parameter competition mechanism, designing compensation step size and performing cross-mutation, thereby improving the accuracy and reliability of the ovarian cancer risk assessment model. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A schematic diagram of an artificial intelligence-based ovarian cancer risk assessment assistance system provided by the present invention;

[0039] Figure 2 Schematic diagram of the ovary feature extraction module;

[0040] Figure 3 Schematic diagram of the module for constructing the ovarian cancer risk assessment model;

[0041] Figure 4 This is a schematic diagram of the model performance improvement module;

[0042] Figure 5 This is a schematic diagram of the process of determining the model performance improvement module.

[0043] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0044] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0045] In the description of the present invention, it should be understood that terms such as "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, they should not be understood as limiting the present invention.

[0046] Example 1, see Figure 1The present invention provides an artificial intelligence-based ovarian cancer risk assessment auxiliary system, which includes a data acquisition module, an ovarian feature extraction module, an ovarian cancer risk assessment model construction module, a model performance improvement module and an ovarian cancer risk assessment auxiliary module;

[0047] The data acquisition module collects clinical data, imaging data, laboratory test data and final disease data of historical patients, and sends the data to the ovarian feature extraction module;

[0048] The ovarian feature extraction module receives data sent by the data acquisition module, extracts ovarian features by detail enhancement, extracts ovarian contours, extracts morphological features, extracts texture features, generates feature vectors, and constructs a model training data set, and sends the data to the ovarian cancer risk assessment model construction module;

[0049] The ovarian cancer risk assessment model construction module receives data sent by the ovarian feature extraction module, constructs an ovarian cancer risk assessment model by setting labels, constructing a double-layer fusion kernel function, designing penalty terms, generating an objective function, and determining optimal conditions, and sends the data to the model performance improvement module;

[0050] The model performance improvement module receives data sent by the ovarian cancer risk assessment model construction module, improves model performance through initialization, parameter performance evaluation, construction of diversity maintenance factors, design of parameter competition mechanism, design of compensation step size, crossover mutation and determination process, and sends the data to the ovarian cancer risk assessment auxiliary module;

[0051] The ovarian cancer risk assessment auxiliary module receives data sent by the model performance improvement module and uses the ovarian cancer risk assessment model to assist in assessing the patient's ovarian cancer risk.

[0052] Example 2, see Figure 1 and Figure 2 This embodiment is based on the above embodiment, and the ovarian feature extraction module specifically includes the following contents:

[0053] Detail enhancement, weighted superposition and Laplacian operator enhancement are performed on the original image. By adaptively adjusting parameters, linear grayscale transformation and normalized detail highlighting are combined, as shown below:

[0054] ;

[0055] Among them, x and y represent the horizontal coordinate index and vertical coordinate index of the pixel point respectively. Indicates that the enhanced image is at pixel point The gray value at 、 and is the detail adaptation parameter, Indicates the original image at pixel point The gray value at and Represent the minimum and maximum grayscale values in the image, respectively. Represents the Laplacian operator of the image;

[0056] To extract the ovarian contour, the horizontal and vertical gradient amplitudes and directions are calculated on the enhanced image. The edge strength and direction information are used to selectively retain the edge pixels that are consistent with the direction of the ovarian contour, thereby extracting the ovarian contour, which is expressed as follows:

[0057] ;

[0058] in, Indicates that the enhanced image is at pixel point The edge strength at represents the symbol of partial derivative, Indicates taking the absolute value, represents the edge direction, represents the inverse tangent function, Indicates the ovarian contour image at pixel point The gray value at represents the cosine function;

[0059] Morphological features were extracted to calculate the area, perimeter, major axis, and minor axis of the contour area, and then the compactness and aspect ratio were calculated to quantitatively describe the regularity and stretching degree of the ovarian shape, as shown below:

[0060] ;

[0061] in, 、 、 、 、 and Respectively represent the compactness, area, perimeter, aspect ratio, major axis length and minor axis length of the ovarian contour image, represents pi;

[0062] Texture feature extraction, based on the gray level co-occurrence matrix, calculates the four texture indices of image contrast, correlation, energy, and homogeneity to characterize the gray level distribution and spatial relationship of the ovarian region, which is expressed as follows:

[0063] ;

[0064] in, 、 、 and Represent the contrast, correlation, energy and homogeneity of the image respectively, i and j represent the horizontal and vertical index of the gray level co-occurrence matrix elements respectively, Represents the gray-level co-occurrence matrix at position The value at Indicates the number of gray levels, and Represents the grayscale mean of the image in the horizontal and vertical directions respectively;

[0065] Feature vector generation, the contrast, correlation, energy, homogeneity, compactness and aspect ratio of the ovarian contour image are combined into the ovarian feature vector, which is expressed as follows:

[0066] ;

[0067] in, represents the ovarian eigenvector;

[0068] Construct a model training data set, add the patient's clinical data, laboratory test data and final disease data to the ovarian feature vector to form a final data vector, create a model training data set, and add the final data vector to the model training data set.

[0069] By performing the above operations, the problems of traditional ovarian feature extraction methods such as detail loss, inaccurate contour extraction, incomplete feature extraction and insufficient data utilization exist. This solution improves the accuracy and comprehensiveness of feature extraction through detail enhancement, accurate contour extraction, comprehensive morphological feature calculation, texture feature calculation, feature integration and data integration, providing richer and more effective data support for model training, thereby improving the accuracy and reliability of ovarian cancer risk assessment.

[0070] Example 3, see Figure 1 and Figure 3 This embodiment is based on the above embodiment, and the construction of the ovarian cancer risk assessment model module specifically includes the following contents:

[0071] Set labels and set the final diseased data as label data for the ovarian cancer risk assessment model;

[0072] Construct a double-layer fusion kernel function by fusing the Gaussian kernel function and the normalization part based on the inverse matrix of the covariance matrix, which is expressed as follows:

[0073] ;

[0074] in, and Represents two data samples in the model training dataset, represents the double-layer fusion kernel function, represents the exponential function with natural parameters as base, represents the bandwidth parameter, Indicates the modulus length, represents the matrix transpose symbol, represents the inverse matrix of the covariance matrix of the model training dataset, Indicates the division by zero factor;

[0075] The penalty term is designed, the spatial smoothness of the kernel function is considered, and the weighted sum of the squared gradient norm is introduced to constrain the complexity of the model, which is expressed as follows:

[0076] ;

[0077] Among them, u represents the data sample variable, represents the weight of the model, represents the penalty term, q and v represent the index of the data sample, M represents the total number of data samples, and Represents data samples respectively and The Lagrange multiplier of and Represents data samples respectively and Tags, Represents the kernel function The gradient with respect to u;

[0078] Generate the objective function. On the basis of maximizing the Lagrangian dual objective, add the quadratic penalty term and the kernel smoothing penalty at the same time to construct the objective function, which is expressed as follows:

[0079] ;

[0080] Where a represents the Lagrange multiplier, It means taking the value of a when the entire expression is at its maximum value. and represents the expansion factor;

[0081] Determine the optimal conditions and, through the Lagrange duality condition, make the gradient of the Lagrange function with respect to the weight and bias zero, and at the same time satisfy the non-negativity of the Lagrange multiplier, the classification interval constraint, and the complementary relaxation, as shown below:

[0082] ;

[0083] in, represents the bias of the model, represents the Lagrangian function About model weights and model bias The gradient of is zero at the optimal solution, making the solution at a stable point in the parameter space; Indicates that for all, means that all Lagrange multipliers are non-negative; Represents data samples Mapping in high-dimensional feature space, Indicates that all samples are correctly classified and the margin is maximized; This means that for each sample, either , or , only the support vectors contribute to the objective function.

[0084] By performing the above operations, the problems of traditional ovarian cancer risk assessment model construction methods, such as a single kernel function, unreasonable penalty terms, incomplete objective function, and unclear optimal conditions, are addressed. This solution improves the model construction quality, classification performance, and stability by constructing a double-layer fusion kernel function, designing innovative penalty terms, generating objective functions, and determining optimal conditions, thereby enhancing the accuracy and reliability of ovarian cancer risk assessment.

[0085] Example 4, see Figure 1 、 Figure 4 and Figure 5 This embodiment is based on the above embodiment, and the model performance improvement module specifically includes the following contents:

[0086] Initialization, determine the search parameters, including bandwidth parameters and scaling factors, and randomly generate initial parameter search points;

[0087] Parameter performance evaluation: set the model accuracy as the parameter performance value and set the parameter performance target value;

[0088] The diversity maintenance factor is constructed by quantifying the breadth of the parameter distribution by taking the average value of the ratio of the range of the search point set in each dimension to the corresponding search width. The diversity maintenance factor is expressed as follows:

[0089] ;

[0090] in, represents the diversity maintenance factor, D represents the total number of dimensions of the parameter search point, d represents the dimension index of the parameter search point, Indicates the search width of the parameter search point in the dth dimension, Indicates the maximum value of the searched parameter point in the dth dimension, Indicates the minimum value of the searched parameter point in the dth dimension;

[0091] Design a parameter competition mechanism, generate a competition probability by scaling the average of the best and worst performance values with a diversity factor, and use this probability to weight the current, best, worst, and average positions to update the parameter point position, as shown below:

[0092] ;

[0093] in, represents the competition probability, represents the optimal parameter performance among all parameter search points, represents the worst parameter performance among all parameter search points, Indicates the parameter position after update, Indicates the parameter position before the update, Indicates the optimal parameter search point among all parameter search points, Indicates the worst parameter search point among all parameter search points, represents the competition adjustment weight, represents the average position of all searched parameter points;

[0094] The compensation step length is designed and calculated based on the ratio of the search point performance standard deviation to the average performance, and the difference between the average position of the initial point and the position of the worst performance point, which is expressed as follows:

[0095] ;

[0096] in, represents the compensation step length, represents the average parameter performance of all searched parameter points, represents the parameter performance standard deviation of all searched parameter points, represents the average position of the initial parameter search points, Indicates the parameter position with the minimum parameter performance among the initial parameter search points;

[0097] Cross-mutation uses Gaussian perturbation and differential mutation strategies on parameter points, and uses compensation steps and randomly selected search points to generate new candidate solutions, which are expressed as follows:

[0098] ;

[0099] in, Indicates the location of the parameter search point after crossover mutation, Indicates the location of the parameter search point before crossover mutation, 、 and represents the mutation weight, represents a random number that obeys the standard normal distribution, r1 represents a random number between 0 and 1, 、 and Indicates the position of a randomly selected parameter search point among all parameter search points;

[0100] Determine the process and set the process of using the parameter competition mechanism to update the position of the parameter point and then perform a cross mutation as a search process; each time a search is performed, calculate the parameter performance value of the parameter search point after the search. If the parameter performance value is greater than the parameter performance standard value, the search stops and the parameter is qualified; otherwise, the parameter is unqualified and the next search is performed.

[0101] By performing the above operations, this solution addresses the problems of unclear parameter performance evaluation, insufficient diversity maintenance, imperfect competition mechanism, unreasonable compensation step size and insufficient crossover mutation in traditional model performance improvement methods. By constructing a diversity maintenance factor, designing a parameter competition mechanism, designing a compensation step size and performing crossover mutation, the efficiency and stability of model performance improvement are improved, thereby improving the accuracy and reliability of the ovarian cancer risk assessment model.

[0102] Example 5, see Figure 1 This embodiment is based on the above embodiment. The ovarian cancer risk assessment auxiliary module collects the user's clinical data, imaging data and laboratory test data, first extracts ovarian features, and then inputs the extracted data into the ovarian cancer risk assessment auxiliary model. The model predicts the user's final disease condition. If the prediction result is disease, the ovarian cancer risk is high risk. If the prediction result is no disease, the ovarian cancer risk is low risk.

[0103] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0104] While the embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that various changes, modifications, substitutions, and alterations can be made to the embodiments without departing from the principles and spirit of the invention.

[0105] The present invention and its embodiments are described above. This description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. An artificial intelligence-based ovarian cancer risk assessment auxiliary system, characterized by: It includes data acquisition module, ovarian feature extraction module, ovarian cancer risk assessment model construction module, model performance improvement module and ovarian cancer risk assessment auxiliary module; The data acquisition module collects clinical data, imaging data, laboratory test data and final disease data of historical patients; The ovarian feature extraction module extracts ovarian features by enhancing details, extracting ovarian contours, extracting morphological features, extracting texture features, generating feature vectors, and constructing a model training data set; The ovarian cancer risk assessment model construction module constructs the ovarian cancer risk assessment model by setting labels, constructing a double-layer fusion kernel function, designing penalty terms, generating an objective function, and determining optimal conditions, specifically including the following: Set labels and set the final diseased data as label data for the ovarian cancer risk assessment model; Construct a double-layer fusion kernel function by fusing the Gaussian kernel function and the normalization part based on the inverse matrix of the covariance matrix; Design the penalty term, consider the spatial smoothness of the kernel function, introduce the weighted sum of the squared gradient norm, and constrain the complexity of the model; Generate the objective function. On the basis of maximizing the Lagrangian dual objective, add the quadratic penalty term and the kernel smoothing penalty to construct the objective function. Determine the optimal conditions, through the Lagrangian dual conditions, make the gradient of the Lagrangian function with respect to weights and biases zero, and at the same time satisfy the non-negativity of the Lagrangian multiplier, the classification interval constraint and the complementary relaxation property; The model performance improvement module improves model performance through initialization, parameter performance evaluation, building a diversity maintenance factor, designing a parameter competition mechanism, designing a compensation step size, crossover mutation, and determining the process. Specifically, it includes the following: Initialization, determine the search parameters, including bandwidth parameters and scaling factors, and randomly generate initial parameter search points; Parameter performance evaluation: set the model accuracy as the parameter performance value and set the parameter performance target value; Construct a diversity maintenance factor, which is obtained by quantifying the breadth of parameter distribution through the average value of the ratio of the range of the search point set in each dimension to the corresponding search width; Design a parameter competition mechanism, generate a competition probability by scaling the average of the best and worst performance values with a diversity factor, and use this probability to weight the current, best, worst, and average positions to update the parameter point position; Design the compensation step size, which is calculated based on the ratio of the search point performance standard deviation to the average performance, and the difference between the average position of the initial point and the position of the worst performance point; Cross-mutation, using Gaussian perturbation and differential mutation strategies on parameter points, using compensation steps and randomly selected other search points to generate new candidate solutions; Determine the process, and set the process of using the parameter competition mechanism to update the position of the parameter point and then perform a cross mutation as a search process; each time a search is performed, calculate the parameter performance value of the parameter search point after the search; if the parameter performance value is greater than the parameter performance standard value, the search stops and the parameter is qualified; otherwise, the parameter is unqualified and the next search is performed; The ovarian cancer risk assessment auxiliary module utilizes an ovarian cancer risk assessment model to assist in assessing a patient's ovarian cancer risk.

2. The artificial intelligence-based ovarian cancer risk assessment auxiliary system according to claim 1, characterized in that: The ovarian feature extraction module specifically includes the following contents: Detail enhancement: weighted superposition and Laplacian operator enhancement are performed on the original image. By adaptively adjusting parameters, linear grayscale transformation is combined with normalized detail highlighting. Extract the ovarian contour by calculating the horizontal and vertical gradient amplitudes and directions on the enhanced image. Utilizing the edge strength and direction information, the edge pixels that are consistent with the direction of the ovarian contour are selectively retained to extract the ovarian contour. Morphological feature extraction: calculating the area, perimeter, major axis, and minor axis of the contour area, and then calculating the compactness and aspect ratio to quantitatively describe the regularity and stretching degree of the ovarian shape; Texture feature extraction: Calculate the image's contrast, correlation, energy, and homogeneity based on the gray-level co-occurrence matrix to characterize the grayscale distribution and spatial relationship of the ovarian region. Feature vector generation, which combines the contrast, correlation, energy, homogeneity, compactness and aspect ratio of the ovarian contour image into an ovarian feature vector; Construct a model training data set, add the patient's clinical data, laboratory test data and final disease data to the ovarian feature vector to form a final data vector, create a model training data set, and add the final data vector to the model training data set.

3. The artificial intelligence-based ovarian cancer risk assessment auxiliary system according to claim 1, characterized in that: The data acquisition module collects clinical data, imaging data, laboratory test data and final disease data of historical patients; the clinical data includes the patient's age, family medical history, number of pregnancies, whether there is abdominal pain and bloating; the imaging data refers to the patient's ultrasound examination image; the laboratory test data includes the test results of tumor markers, blood routine, liver function, kidney function and hormone levels; the final disease data refers to whether the patient has ovarian cancer; The ovarian cancer risk assessment auxiliary module collects the user's clinical data, imaging data and laboratory test data, first extracts ovarian features, and then inputs the extracted data into the ovarian cancer risk assessment auxiliary model. The model predicts the user's final disease status. If the prediction result is that the user is ill, the ovarian cancer risk is high risk. If the prediction result is that the user is not ill, the ovarian cancer risk is low risk.

Images