Artificial intelligence-based pelvic floor ultrasound anomaly early warning system
Patent Information
- Application Number
- CN202411739711.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-29
- Publication Date
- 2026-10-23
AI Technical Summary
[0003]针对上述情况，为克服现有技术的缺陷，本发明提供了基于人工智能的盆底超声异常预警系统，针对传统盆底超声图像处理与分析存在处理效率低下、特征提取不充分以及易受噪声干扰的问题，本方案通过数据标注、定义滤波权重、梯度自适应滤波、图像分解、分量增强变换、图像重建和多尺度特征提取，保留了图像的全局结构信息，突出了图像的局部细节特征，进一步提高了特征提取的全面性和准确性，削弱噪声的影响，使得最终的重建图像更加清晰、准确；针对传统盆底超声异常检测模型存在特征提取能力不足、模型泛化性能差以及计算复杂度高的问题，本方案通过构建模型架构、设计卷积层、归一化层、池化层和分类层，更有效地捕捉盆底超声图像中的细微特征，提高了模型对异常区域的识别能力，使得模型在训练过程中更容易收敛，提高了模型的分类准确性；针对传统盆底超声异常检测模型的全局优化存在搜索效率低、易陷入局部最优解以及参数调整依赖人工经验的问题，本方案通过全局优化准备、定义初始化权重、生成起源参数点、设计螺旋步长、计算螺旋角、设计螺旋更新函数和设计迭代规则，不断调整步长和角度，螺旋更新函数能够逐渐逼近全局最优解，从而提高了搜索效率，增加了搜索的多样性和灵活性，降低了陷入局部最优解的风险，减少了人工干预和依赖，提高了全局优化的准确性和可靠性
[0065]（1）针对传统盆底超声图像处理与分析存在处理效率低下、特征提取不充分以及易受噪声干扰的问题，本方案通过数据标注、定义滤波权重、梯度自适应滤波、图像分解、分量增强变换、图像重建和多尺度特征提取，保留了图像的全局结构信息，突出了图像的局部细节特征，进一步提高了特征提取的全面性和准确性，削弱噪声的影响，使得最终的重建图像更加清晰、准确。
Smart Images
Figure CN122966496A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of medical early warning technology, specifically to an artificial intelligence-based pelvic floor ultrasound abnormality early warning system. Background Technology
[0002] The AI-based pelvic floor ultrasound anomaly early warning system utilizes artificial intelligence and ultrasound imaging technologies to perform real-time, precise analysis and evaluation of pelvic floor tissue structures, identify subtle abnormal changes, and issue timely warning signals. However, traditional pelvic floor ultrasound image processing and analysis suffers from low processing efficiency, insufficient feature extraction, and susceptibility to noise interference; traditional pelvic floor ultrasound anomaly detection models suffer from insufficient feature extraction capabilities, poor model generalization performance, and high computational complexity; and the global optimization of traditional pelvic floor ultrasound anomaly detection models suffers from low search efficiency, susceptibility to getting trapped in local optima, and reliance on human experience for parameter adjustment. Summary of the Invention
[0003] To address the above issues and overcome the shortcomings of existing technologies, this invention provides an artificial intelligence-based pelvic floor ultrasound anomaly early warning system. Addressing the problems of low processing efficiency, insufficient feature extraction, and susceptibility to noise interference in traditional pelvic floor ultrasound image processing and analysis, this solution preserves the global structural information of the image and highlights local detail features through data annotation, defining filter weights, gradient adaptive filtering, image decomposition, component enhancement transformation, image reconstruction, and multi-scale feature extraction. This further improves the comprehensiveness and accuracy of feature extraction, reduces the impact of noise, and makes the final reconstructed image clearer and more accurate. Furthermore, addressing the problems of insufficient feature extraction capability, poor model generalization performance, and high computational complexity in traditional pelvic floor ultrasound anomaly detection models, this solution constructs a model architecture, designs convolutional layers, normalization layers, and... Pooling and classification layers more effectively capture subtle features in pelvic floor ultrasound images, improving the model's ability to identify abnormal regions and making the model converge more easily during training, thus improving classification accuracy. Addressing the issues of low search efficiency, susceptibility to local optima, and reliance on human experience in global optimization of traditional pelvic floor ultrasound anomaly detection models, this solution employs global optimization preparation, defining initial weights, generating origin parameter points, designing spiral step size, calculating spiral angle, designing spiral update function, and designing iteration rules. By continuously adjusting the step size and angle, the spiral update function gradually approaches the global optimum, thereby improving search efficiency, increasing search diversity and flexibility, reducing the risk of getting trapped in local optima, minimizing human intervention and reliance, and improving the accuracy and reliability of global optimization.
[0004] The technical solution adopted by the present invention is as follows: The pelvic floor ultrasound abnormality early warning system based on artificial intelligence provided by the present invention includes a data acquisition module, a data preprocessing module, a pelvic floor ultrasound abnormality detection model construction module, a global optimization module, and an abnormality early warning module;
[0005] The data acquisition module collects historical pelvic floor ultrasound images and pelvic floor ultrasound status data from patients; the pelvic floor ultrasound status data includes normal and abnormal states.
[0006] The data preprocessing module preprocesses the data through data labeling, defining filter weights, gradient adaptive filtering, image decomposition, component enhancement transformation, image reconstruction, and multi-scale feature extraction.
[0007] The module for constructing a pelvic floor ultrasound anomaly detection model builds a pelvic floor ultrasound anomaly detection model by constructing a model architecture and designing convolutional layers, normalization layers, pooling layers, and classification layers.
[0008] The global optimization module optimizes the model by preparing for global optimization, defining initial weights, generating origin parameter points, designing spiral step size, calculating spiral angle, designing spiral update function, and designing iteration rules.
[0009] The abnormality warning module uses the pelvic floor ultrasound abnormality detection model to output the status of the patient's pelvic floor ultrasound image. When an abnormal status is detected, an early warning signal is issued.
[0010] Furthermore, the data preprocessing module specifically includes the following:
[0011] Data annotation: Each pelvic floor ultrasound image was annotated using pelvic floor ultrasound status data;
[0012] Define the filter weights as follows:
[0013] ;
[0014] Where m and n represent the x-coordinate index and y-coordinate index of the filter window, respectively, and x and y represent the x-coordinate index and y-coordinate index of the image pixel, respectively. Indicated by image position Centered on, the filter window is located at position The filter weights at the point, This indicates the location of the original image with the filter window as the local region. Local pixel standard deviation at that location Here, e represents the smoothing factor, and e represents the natural constant. Indicates the gradient image at position The value at that location, This indicates the gradient image at position with the filter window as the local region. Local pixel mean at that location The pixel standard deviation of the gradient image. Indicates taking the absolute value;
[0015] Gradient adaptive filtering is represented as follows:
[0016] ;
[0017] in, Indicates the position of the filtered image. Pixel value at that location, Indicates the original image at position Pixel value at that location, Indicates the gradient image at position The value at that location, This indicates the location of the original image with the filter window as the local region. The local pixel mean at that location;
[0018] Image decomposition, represented as follows:
[0019] ;
[0020] in, This represents the approximate image obtained from the decomposition. This indicates a downsampling operation. This represents the detailed image after decomposition. Indicates an upsampling operation;
[0021] The component enhancement transform is represented as follows:
[0022] ;
[0023] in, This represents the approximate image after the enhancement transformation. , , and Indicates the transformation factor. This represents the detailed image after enhancement transformation. Indicates taking the logarithm. Represents a symbolic function;
[0024] Image reconstruction involves performing the inverse process of image decomposition on the enhanced image to reconstruct the image.
[0025] Design a feature extraction function, expressed as follows:
[0026] ;
[0027] in, This represents the feature extraction function, where 'a' represents the feature scale parameter. and Indicates the translation parameter. and Let these represent the cosine and sine functions, respectively. Indicates the direction parameter. Indicates the fluctuation parameter. represents the square root, and i represents the imaginary unit;
[0028] Multi-scale feature extraction is represented as follows:
[0029] ;
[0030] in, This represents the reconstructed image. This represents the feature coefficients extracted from the reconstructed image.
[0031] Furthermore, the module for constructing a pelvic floor ultrasound abnormality detection model specifically includes the following:
[0032] Construct the model architecture, which includes convolutional layers, normalization layers, pooling layers, fully connected layers, and classification layers;
[0033] The convolutional layer is designed as follows:
[0034] ;
[0035] in, and These represent the row and column indices of the feature map, respectively. This indicates the position of the output feature map of the convolutional layer. The value at that location, and Let represent the height index and width index of the convolution kernel, respectively, and let c represent the channel index of the input feature map. This indicates that the c-th channel of the input feature map is at position The value at that location, This indicates that the convolution kernel of the c-th channel is located at position... The value at that location, Indicates the bias of the convolutional layer;
[0036] The normalization layer is designed as follows:
[0037] ;
[0038] in, This represents the output of the normalization layer. and These represent the weights and biases of the normalization layer, respectively. This represents the input to the normalization layer. This represents the mean of the input to the normalization layer. This represents the standard deviation of the input to the normalization layer. Indicates removing zero parameters;
[0039] The pooling layer is designed as follows:
[0040] ;
[0041] in, This indicates the position of the output feature map of the pooling layer. The value at that location, max indicates taking the maximum value. and These represent the x-coordinate index and y-coordinate index of the pooling window, respectively. Indicates a pooled window. Indicates that it belongs to the symbol. and These represent the width and height of the pooling window, respectively. This indicates that the input feature map of the pooling layer is at the location The value at;
[0042] The classification layers are designed as follows:
[0043] ;
[0044] in, This represents the output of the classification layer. Represents the normalized exponential function, , and These represent the weights, inputs, and biases of the classification layer, respectively.
[0045] Furthermore, the global optimization module specifically includes the following:
[0046] Global optimization preparation involves determining optimization parameters, including the weights, biases, and width and height of the pooling window for the pelvic floor ultrasound anomaly detection model; setting the model's accuracy as the performance intensity of individual parameters and defining the maximum number of parameter searches.
[0047] Define the initial weights as follows:
[0048] ;
[0049] Where f represents the number of times the origin parameter point is generated. Let F represent the initial weights when the origin parameter point is generated for the f-th time, and F represent the maximum number of times the origin parameter point can be generated. Represents pi (π). Indicates the upper bound of the parameter search space. Indicates the lower bound of the parameter search space. This represents a parameter position that is randomly generated within the parameter search space;
[0050] The origin parameter point is generated as follows:
[0051] ;
[0052] in, This indicates the position of the origin parameter point generated in the (f+1)th generation. This indicates the position of the origin parameter point generated in the f-th generation. This represents a random number that follows a standard normal distribution.
[0053] The spiral step size is designed as follows:
[0054] ;
[0055] Where t represents the current number of searches for the parameter. This represents the spiral step size during the t-th parameter search. This indicates the position where the parameter performance is at its highest. This indicates the position where the parameter performance strength is lowest. This represents a random number whose value is between 0 and 1. This represents the average position of the generated origin parameter points. This indicates the maximum number of parameter searches;
[0056] The helix angle is calculated as follows:
[0057] ;
[0058] in, This represents the spiral angle during the t-th parameter search. This represents a random number whose value is between 0 and 1.
[0059] Design a spiral update function, represented as follows:
[0060] ;
[0061] in, This represents the parameter position obtained through the spiral update function during the (t+1)th parameter search. This represents the parameter position obtained through the spiral update function during the t-th parameter search. This represents a random number whose value is between 0 and 1. This represents the average position of all parameter points obtained in the t-th parameter search;
[0062] The design iterative rules set a threshold value for parameter performance strength. A spiral update function is used to search for parameters at the origin parameter point. During the search, if the parameter performance strength at a parameter position is greater than the threshold value, the search is stopped, and the parameter at the position with the highest parameter performance strength is set as the model parameter. If the search count reaches the maximum parameter search count, the search is restarted. Otherwise, the search count is incremented by one, and the search continues.
[0063] Furthermore, the abnormality warning module acquires the patient's pelvic floor ultrasound image, inputs the image into the pelvic floor ultrasound abnormality detection model, and the model outputs the status of the patient's pelvic floor ultrasound image. When the status of the pelvic floor ultrasound image is abnormal, a warning signal is issued.
[0064] The beneficial effects achieved by adopting the above solution are as follows:
[0065] (1) In view of the problems of low processing efficiency, insufficient feature extraction and susceptibility to noise interference in traditional pelvic floor ultrasound image processing and analysis, this scheme preserves the global structural information of the image and highlights the local details of the image by data annotation, defining filter weights, gradient adaptive filtering, image decomposition, component enhancement transformation, image reconstruction and multi-scale feature extraction, thereby improving the comprehensiveness and accuracy of feature extraction, weakening the influence of noise, and making the final reconstructed image clearer and more accurate.
[0066] (2) In view of the problems of insufficient feature extraction ability, poor generalization performance and high computational complexity of traditional pelvic floor ultrasound anomaly detection models, this solution constructs a model architecture and designs convolutional layer, normalization layer, pooling layer and classification layer to more effectively capture subtle features in pelvic floor ultrasound images, improve the model's ability to identify abnormal areas, make the model more likely to converge during training, and improve the model's classification accuracy.
[0067] (3) To address the problems of low search efficiency, easy getting trapped in local optima, and reliance on human experience in the global optimization of traditional pelvic floor ultrasound anomaly detection models, this scheme continuously adjusts the step size and angle by preparing for global optimization, defining initial weights, generating origin parameter points, designing spiral step size, calculating spiral angle, designing spiral update function, and designing iteration rules. The spiral update function can gradually approach the global optimum, thereby improving search efficiency, increasing the diversity and flexibility of the search, reducing the risk of getting trapped in local optima, reducing human intervention and reliance, and improving the accuracy and reliability of global optimization. Attached Figure Description
[0068] Figure 1 A schematic diagram of the pelvic floor ultrasound abnormality early warning system based on artificial intelligence provided by the present invention;
[0069] Figure 2 This is a schematic diagram of the data preprocessing module;
[0070] Figure 3 A schematic diagram of a module for constructing a pelvic floor ultrasound anomaly detection model;
[0071] Figure 4 This is a schematic diagram of the global optimization module.
[0072] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation
[0073] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0074] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.
[0075] Example 1, see Figure 1 The artificial intelligence-based pelvic floor ultrasound abnormality early warning system provided by the present invention includes a data acquisition module, a data preprocessing module, a pelvic floor ultrasound abnormality detection model construction module, a global optimization module, and an abnormality early warning module.
[0076] The data acquisition module acquires pelvic floor ultrasound images and pelvic floor ultrasound status data from historical patients; the pelvic floor ultrasound status data includes normal and abnormal states, and the data is sent to the data preprocessing module.
[0077] The data preprocessing module preprocesses the data through data labeling, defining filter weights, gradient adaptive filtering, image decomposition, component enhancement transformation, image reconstruction, and multi-scale feature extraction, and then sends the data to the module for constructing a pelvic floor ultrasound anomaly detection model.
[0078] The module for constructing a pelvic floor ultrasound anomaly detection model builds the model by constructing a model architecture, designing convolutional layers, normalization layers, pooling layers, and classification layers, and sends the data to the global optimization module.
[0079] The global optimization module optimizes the model by preparing for global optimization, defining initial weights, generating origin parameter points, designing spiral step size, calculating spiral angle, designing spiral update function, and designing iteration rules, and sends the data to the anomaly warning module.
[0080] The abnormality warning module uses the pelvic floor ultrasound abnormality detection model to output the status of the patient's pelvic floor ultrasound image. When an abnormal status is detected, an early warning signal is issued.
[0081] Example 2, see Figure 1 and Figure 2 This embodiment is based on the above embodiment, and the data preprocessing module specifically includes the following:
[0082] Data annotation: Each pelvic floor ultrasound image was annotated using pelvic floor ultrasound status data;
[0083] Define the filter weights as follows:
[0084] ;
[0085] Where m and n represent the x-coordinate index and y-coordinate index of the filter window, respectively, and x and y represent the x-coordinate index and y-coordinate index of the image pixel, respectively. Indicated by image position Centered on, the filter window is located at position The filter weights at the point, This indicates the location of the original image with the filter window as the local region. Local pixel standard deviation at that location Here, e represents the smoothing factor, and e represents the natural constant. Indicates the gradient image at position The value at that location, This indicates the gradient image at position with the filter window as the local region. Local pixel mean at that location The pixel standard deviation of the gradient image. Indicates taking the absolute value;
[0086] Gradient adaptive filtering is represented as follows:
[0087] ;
[0088] in, Indicates the position of the filtered image. Pixel value at that location, Indicates the original image at position Pixel value at that location, Indicates the gradient image at position The value at that location, This indicates the location of the original image with the filter window as the local region. The local pixel mean at that location;
[0089] Image decomposition, represented as follows:
[0090] ;
[0091] in, This represents the approximate image obtained from the decomposition. This indicates a downsampling operation. This represents the detailed image after decomposition. Indicates an upsampling operation;
[0092] The component enhancement transform is represented as follows:
[0093] ;
[0094] in, This represents the approximate image after the enhancement transformation. , , and Indicates the transformation factor. This represents the detailed image after enhancement transformation. Indicates taking the logarithm. Represents a symbolic function;
[0095] Image reconstruction involves performing the inverse process of image decomposition on the enhanced image to reconstruct the image.
[0096] Design a feature extraction function, expressed as follows:
[0097] ;
[0098] in, This represents the feature extraction function, where 'a' represents the feature scale parameter. and Indicates the translation parameter. and Let these represent the cosine and sine functions, respectively. Indicates the direction parameter. Indicates the fluctuation parameter. represents the square root, and i represents the imaginary unit;
[0099] Multi-scale feature extraction is represented as follows:
[0100] ;
[0101] in, This represents the reconstructed image. This represents the feature coefficients extracted from the reconstructed image.
[0102] By performing the above operations, this solution addresses the problems of low processing efficiency, insufficient feature extraction, and susceptibility to noise interference in traditional pelvic floor ultrasound image processing and analysis. Through data annotation, defining filter weights, gradient adaptive filtering, image decomposition, component enhancement transformation, image reconstruction, and multi-scale feature extraction, it preserves the global structural information of the image, highlights the local details of the image, further improves the comprehensiveness and accuracy of feature extraction, reduces the impact of noise, and makes the final reconstructed image clearer and more accurate.
[0103] Example 3, see Figure 1 and Figure 3 This embodiment is based on the above embodiment, and the module for constructing the pelvic floor ultrasound abnormality detection model specifically includes the following:
[0104] Construct the model architecture, which includes convolutional layers, normalization layers, pooling layers, fully connected layers, and classification layers;
[0105] The convolutional layer is designed as follows:
[0106] ;
[0107] in, and These represent the row and column indices of the feature map, respectively. This indicates the position of the output feature map of the convolutional layer. The value at that location, and Let represent the height index and width index of the convolution kernel, respectively, and let c represent the channel index of the input feature map. This indicates that the c-th channel of the input feature map is at position The value at that location, This indicates that the convolution kernel of the c-th channel is located at position The value at that location, Indicates the bias of the convolutional layer;
[0108] The normalization layer is designed as follows:
[0109] ;
[0110] in, This represents the output of the normalization layer. and These represent the weights and biases of the normalization layer, respectively. This represents the input to the normalization layer. This represents the mean of the input to the normalization layer. This represents the standard deviation of the input to the normalization layer. Indicates removing zero parameters;
[0111] The pooling layer is designed as follows:
[0112] ;
[0113] in, This indicates the position of the output feature map of the pooling layer. The value at that location, max indicates taking the maximum value. and These represent the x-coordinate index and y-coordinate index of the pooling window, respectively. Indicates a pooled window. Indicates that it belongs to the symbol. and These represent the width and height of the pooling window, respectively. This indicates that the input feature map of the pooling layer is at the location The value at;
[0114] The classification layers are designed as follows:
[0115] ;
[0116] in, This represents the output of the classification layer. Represents the normalized exponential function, , and These represent the weights, inputs, and biases of the classification layer, respectively.
[0117] By performing the above operations, this solution addresses the problems of insufficient feature extraction capability, poor model generalization performance, and high computational complexity in traditional pelvic floor ultrasound anomaly detection models. It constructs a model architecture and designs convolutional layers, normalization layers, pooling layers, and classification layers to more effectively capture subtle features in pelvic floor ultrasound images, improves the model's ability to identify abnormal regions, makes the model more likely to converge during training, and improves the model's classification accuracy.
[0118] Example 4, see Figure 1 and Figure 4 This embodiment is based on the above embodiment, and the global optimization module specifically includes the following:
[0119] Global optimization preparation involves determining optimization parameters, including the weights, biases, and width and height of the pooling window for the pelvic floor ultrasound anomaly detection model; setting the model's accuracy as the performance intensity of individual parameters and defining the maximum number of parameter searches.
[0120] Define the initial weights as follows:
[0121] ;
[0122] Where f represents the number of times the origin parameter point is generated. Let F represent the initial weights for the f-th generation of the origin parameter point, and F represent the maximum number of times the origin parameter point can be generated. Represents pi (π). Indicates the upper bound of the parameter search space. Indicates the lower bound of the parameter search space. This represents a parameter position that is randomly generated within the parameter search space;
[0123] The origin parameter point is generated as follows:
[0124] ;
[0125] in, This indicates the position of the origin parameter point generated in the (f+1)th generation. This indicates the position of the origin parameter point generated in the f-th generation. This represents a random number that follows a standard normal distribution.
[0126] The spiral step size is designed as follows:
[0127] ;
[0128] Where t represents the current number of searches for the parameter. This represents the spiral step size during the t-th parameter search. This indicates the position where the parameter performance is at its highest. This indicates the position where the parameter performance strength is lowest at this time. This represents a random number whose value is between 0 and 1. This indicates the average location of the generated origin parameter points. This indicates the maximum number of parameter searches;
[0129] The helix angle is calculated as follows:
[0130] ;
[0131] in, This represents the spiral angle during the t-th parameter search. This represents a random number whose value is between 0 and 1.
[0132] Design a spiral update function, represented as follows:
[0133] ;
[0134] in, This represents the parameter position obtained through the spiral update function during the (t+1)th parameter search. This represents the parameter position obtained through the spiral update function during the t-th parameter search. This represents a random number whose value is between 0 and 1. This represents the average position of all parameter points obtained in the t-th parameter search;
[0135] The design iterative rules set a threshold value for parameter performance strength. A spiral update function is used to search for parameters at the origin parameter point. During the search, if the parameter performance strength at a parameter position is greater than the threshold value, the search is stopped, and the parameter at the position with the highest parameter performance strength is set as the model parameter. If the search count reaches the maximum parameter search count, the search is restarted. Otherwise, the search count is incremented by one, and the search continues.
[0136] By performing the above operations, this scheme addresses the problems of low search efficiency, susceptibility to local optima, and reliance on human experience in the global optimization of traditional pelvic floor ultrasound anomaly detection models. It improves search efficiency, increases search diversity and flexibility, reduces the risk of getting trapped in local optima, minimizes human intervention and reliance on human experience, and enhances the accuracy and reliability of global optimization. This is achieved through global optimization preparation, defining initial weights, generating origin parameter points, designing spiral step sizes, calculating spiral angles, designing spiral update functions, and designing iteration rules. By continuously adjusting the step size and angle, the spiral update function gradually approaches the global optimum, thereby improving search efficiency, increasing search diversity and flexibility, reducing the risk of getting trapped in local optima, minimizing human intervention and reliance on human experience, and improving the accuracy and reliability of global optimization.
[0137] Example 5, see Figure 1 This embodiment is based on the above embodiment. The abnormality warning module collects the patient's pelvic floor ultrasound image, inputs the image into the pelvic floor ultrasound abnormality detection model, and the model outputs the status of the patient's pelvic floor ultrasound image. When the status of the pelvic floor ultrasound image is abnormal, a warning signal is issued.
[0138] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.
[0139] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.
[0140] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.
Claims
1. An artificial intelligence-based pelvic floor ultrasound abnormality early warning system, characterized in that: It includes a data acquisition module, a data preprocessing module, a pelvic floor ultrasound anomaly detection model construction module, a global optimization module, and an anomaly early warning module; The data acquisition module collects historical pelvic floor ultrasound images and pelvic floor ultrasound status data from patients; the pelvic floor ultrasound status data includes normal and abnormal states. The data preprocessing module preprocesses the data through data labeling, defining filter weights, gradient adaptive filtering, image decomposition, component enhancement transformation, image reconstruction, and multi-scale feature extraction. The module for constructing a pelvic floor ultrasound anomaly detection model builds a pelvic floor ultrasound anomaly detection model by constructing a model architecture and designing convolutional layers, normalization layers, pooling layers, and classification layers. The global optimization module optimizes the model by preparing for global optimization, defining initial weights, generating origin parameter points, designing spiral step size, calculating spiral angle, designing spiral update function, and designing iteration rules. The abnormality warning module uses the pelvic floor ultrasound abnormality detection model to output the status of the patient's pelvic floor ultrasound image. When an abnormal status is detected, an early warning signal is issued.
2. The pelvic floor ultrasound abnormality early warning system based on artificial intelligence according to claim 1, characterized in that: The data preprocessing module specifically includes the following: Data annotation: Each pelvic floor ultrasound image was annotated using pelvic floor ultrasound status data; Define the filter weights as follows: ； Where m and n represent the x-coordinate index and y-coordinate index of the filter window, respectively, and x and y represent the x-coordinate index and y-coordinate index of the image pixel, respectively. Indicated by image position Centered on, the filter window is located at position The filter weights at the point, This indicates the location of the original image with the filter window as the local region. Local pixel standard deviation at that location Here, e represents the smoothing factor, and e represents the natural constant. Indicates the gradient image at position The value at that location, This indicates the gradient image at position with the filter window as the local region. Local pixel mean at that location The pixel standard deviation of the gradient image. Indicates taking the absolute value; Gradient adaptive filtering is represented as follows: ； in, Indicates the position of the filtered image. Pixel value at that location, Indicates the original image at position Pixel value at that location, Indicates the gradient image at position The value at that location, This indicates the location of the original image with the filter window as the local region. Local pixel mean at; Image decomposition, represented as follows: ； in, This represents the approximate image obtained from the decomposition. This indicates a downsampling operation. This represents the detailed image after decomposition. Indicates an upsampling operation; The component enhancement transform is represented as follows: ； in, This represents the approximate image after the enhancement transformation. , , and Indicates the transformation factor. This represents the detailed image after enhancement transformation. Indicates taking the logarithm. Represents a symbolic function; Image reconstruction involves performing the inverse process of image decomposition on the enhanced image to reconstruct the image. Design a feature extraction function, expressed as follows: ； in, This represents the feature extraction function, where 'a' represents the feature scale parameter. and Indicates the translation parameter. and Let these represent the cosine and sine functions, respectively. Indicates the direction parameter. Indicates the fluctuation parameter. represents the square root, and i represents the imaginary unit; Multi-scale feature extraction is represented as follows: ； in, This represents the reconstructed image. This represents the feature coefficients extracted from the reconstructed image.
3. The pelvic floor ultrasound abnormality early warning system based on artificial intelligence according to claim 1, characterized in that: The module for constructing a pelvic floor ultrasound abnormality detection model specifically includes the following: Construct the model architecture, which includes convolutional layers, normalization layers, pooling layers, fully connected layers, and classification layers; The convolutional layer is designed as follows: ； in, and These represent the row and column indices of the feature map, respectively. This indicates the position of the output feature map of the convolutional layer. The value at that location, and Let represent the height index and width index of the convolution kernel, respectively, and let c represent the channel index of the input feature map. This indicates that the c-th channel of the input feature map is at position The value at that location, This indicates that the convolution kernel of the c-th channel is located at position... The value at that location, Indicates the bias of the convolutional layer; The normalization layer is designed as follows: ； in, This represents the output of the normalization layer. and These represent the weights and biases of the normalization layer, respectively. This represents the input to the normalization layer. This represents the mean of the input to the normalization layer. This represents the standard deviation of the input to the normalization layer. Indicates removing zero parameters; The pooling layer is designed as follows: ； in, This indicates the position of the output feature map of the pooling layer. The value at that location, max indicates taking the maximum value. and These represent the x-coordinate index and y-coordinate index of the pooling window, respectively. Indicates a pooled window. Indicates that it belongs to the symbol. and These represent the width and height of the pooling window, respectively. This indicates that the input feature map of the pooling layer is at the location The value at; The classification layers are designed as follows: ； in, This represents the output of the classification layer. Represents the normalized exponential function, , and These represent the weights, inputs, and biases of the classification layer, respectively.
4. The pelvic floor ultrasound abnormality early warning system based on artificial intelligence according to claim 1, characterized in that: The global optimization module specifically includes the following: Global optimization preparation involves determining optimization parameters, including the weights, biases, and width and height of the pooling window for the pelvic floor ultrasound anomaly detection model; setting the model's accuracy as the performance intensity of individual parameters and defining the maximum number of parameter searches. Define the initial weights as follows: ； Where f represents the number of times the origin parameter point is generated. Let F represent the initial weights for the f-th generation of the origin parameter point, and F represent the maximum number of times the origin parameter point can be generated. Represents pi (π). Indicates the upper bound of the parameter search space. Indicates the lower bound of the parameter search space. This represents a parameter position that is randomly generated within the parameter search space; The origin parameter point is generated as follows: ； in, This indicates the position of the origin parameter point generated in the (f+1)th generation. This indicates the position of the origin parameter point generated in the f-th generation. This represents a random number that follows a standard normal distribution. The spiral step size is designed as follows: ； Where t represents the current number of searches for the parameter. This represents the spiral step size during the t-th parameter search. This indicates the position where the parameter performance is at its highest. This indicates the position where the parameter performance strength is lowest. This represents a random number whose value is between 0 and 1. This represents the average position of the generated origin parameter points. This indicates the maximum number of parameter searches; The helix angle is calculated as follows: ； in, This represents the spiral angle during the t-th parameter search. This represents a random number whose value is between 0 and 1. Design a spiral update function, represented as follows: ； in, This represents the parameter position obtained through the spiral update function during the (t+1)th parameter search. This represents the parameter position obtained through the spiral update function during the t-th parameter search. This represents a random number whose value is between 0 and 1. This represents the average position of all parameter points obtained in the t-th parameter search; The design iterative rules set a threshold value for parameter performance strength. A spiral update function is used to search for parameters at the origin parameter point. During the search, if the parameter performance strength at a parameter position is greater than the threshold value, the search is stopped, and the parameter at the position with the highest parameter performance strength is set as the model parameter. If the search count reaches the maximum parameter search count, the search is restarted. Otherwise, the search count is incremented by one, and the search continues.
5. The pelvic floor ultrasound abnormality early warning system based on artificial intelligence according to claim 1, characterized in that: The abnormality warning module acquires the patient's pelvic floor ultrasound images, inputs the images into the pelvic floor ultrasound abnormality detection model, and the model outputs the status of the patient's pelvic floor ultrasound images. When the status of the pelvic floor ultrasound images is abnormal, a warning signal is issued.