A three-dimensional reconstruction and lesion detection system for breast ultrasound images

Abstract

The present invention discloses a three-dimensional reconstruction and lesion detection system for breast ultrasound images, including an ultrasound data acquisition unit, a breast ultrasound image preprocessing module, a three-dimensional reconstruction module, a lesion detection and classification module, a data storage and management module, and a report generation and display interaction module. The ultrasound data acquisition unit is used to acquire breast ultrasound images, the breast ultrasound image preprocessing module is used for image preprocessing, the three-dimensional reconstruction module is used for three-dimensional reconstruction of images, the lesion detection and classification module is used to detect and classify breast lesions, the data storage and management module is used to store and manage data and detection reports, and the report generation and display interaction module is used to provide reports and interactive interfaces. The three-dimensional reconstruction and lesion detection system for breast ultrasound images according to the present invention proposes a three-dimensional modeling algorithm for breast ultrasound images based on tomography for three-dimensional reconstruction of breast images, and proposes a breast lesion detection algorithm based on a fuzzy neural network for detecting and identifying breast lesions.

Description

Technical Field

[0001] The present invention relates to the fields of tomography, three-dimensional reconstruction, and fuzzy neural networks, and specifically provides a three-dimensional reconstruction and lesion detection system for breast ultrasound images. Background Art

[0002] Tomography technology and three-dimensional reconstruction technology are imaging techniques that use multi-angle scanning to obtain internal structure information of an object, aiming to solve the problem that traditional two-dimensional images in breast ultrasound images are difficult to comprehensively display the deep structure of breast tissue. Through tomography technology and three-dimensional reconstruction technology, multi-level information inside the breast is collected from different angles, and a three-dimensional breast image is reconstructed, enabling the system to intuitively and clearly reconstruct the internal organizational structure of the breast. Tomography technology provides the core means of image acquisition and reconstruction, which can greatly improve the resolution and accuracy of internal breast images, providing a more reliable basis for subsequent lesion identification and precise treatment.

[0003] Fuzzy neural network technology is an intelligent algorithm that combines fuzzy logic and neural networks, aiming to solve the problems of insufficient accuracy and adaptability of traditional algorithms in processing breast lesion detection. Since the lesion areas in breast ultrasound images vary greatly in size, shape, and density, the fuzzy neural network processes different lesion features through fuzzy processing to better handle complex data and improve the robustness and sensitivity of lesion detection, enabling the system to automatically identify lesion tissues, achieve accurate judgment of lesion types, reduce the subjective influence of doctors, and thus improve the accuracy and consistency of diagnosis.

[0004] However, an existing three-dimensional reconstruction and lesion detection system for breast ultrasound images is difficult to display the deep three-dimensional structure of the breast, resulting in missed and misjudged lesions. At the same time, the lack of three-dimensional reconstruction function limits the three-dimensional observation and analysis of breast tissue. In terms of lesion detection, traditional systems are difficult to accurately handle complex breast tissue differences, often have low sensitivity to lesion detection, are easily interfered by noise, and lead to the risks of misdiagnosis and missed diagnosis. Summary of the Invention

[0005] The purpose of the present invention is to provide a three-dimensional reconstruction and lesion detection system for breast ultrasound images to solve the problems existing in the existing three-dimensional reconstruction and lesion detection system for breast ultrasound images as mentioned in the above background art, namely, the system is difficult to display the deep three-dimensional structure of the breast, resulting in missed and misjudged lesions. At the same time, the lack of three-dimensional reconstruction function in the system limits the three-dimensional observation and analysis of breast tissue. In terms of lesion detection, traditional systems are difficult to accurately handle complex breast tissue differences, often have low sensitivity to lesion detection, are easily interfered by noise, and lead to the risks of misdiagnosis and missed diagnosis.

[0006] To achieve the above object, the present invention provides the following technical solutions: A three-dimensional reconstruction and lesion detection system for breast ultrasound images, comprising an ultrasound data acquisition unit, a breast ultrasound image preprocessing module, a three-dimensional reconstruction module, a lesion detection and classification module, a data storage and management module, and a report generation and display interaction module, characterized in that: The ultrasound data acquisition unit is used to collect breast ultrasound image data, ensuring the acquisition of high-quality original image information and providing a basis for subsequent processing. The breast ultrasound image preprocessing module is used to perform preprocessing operations such as filtering and noise reduction, contrast enhancement, and image edge preservation on the acquired ultrasound images, ensuring image clarity and quality and guaranteeing the accuracy of subsequent lesion detection. The three-dimensional reconstruction module includes an image registration unit and a three-dimensional modeling unit. The image registration unit is used to spatially align the ultrasound images obtained from different perspectives and time points, ensuring consistency and comparability between images and providing accurate image data for three-dimensional modeling. The three-dimensional modeling unit proposes a three-dimensional modeling algorithm for breast ultrasound images based on tomography to construct a three-dimensional model according to the registered two-dimensional ultrasound images, ensuring that the three-dimensional structures of breast tissues and lesions can be accurately presented and analyzed. The lesion detection and classification module includes a feature extraction unit and a classification and recognition unit. The feature extraction unit is used to extract key information from breast ultrasound images, including breast morphology, texture, and boundary features, ensuring rich local and global feature data for classification. The classification and recognition unit proposes a breast lesion detection algorithm based on a fuzzy neural network to classify and recognize breast lesions according to the extracted features, ensuring accurate detection and effective differentiation of breast lesions by the system. The data storage and management module is used to store and manage the ultrasound image data and detection reports of patients, ensuring the security, reliability, and efficient access of the data. The report generation and display interaction module is used to generate detection reports and provide a user interaction interface, enabling users to view reports and images in real time, ensuring the accurate transmission of detection results and the convenience of user operations.

[0007] Preferably, the ultrasound data acquisition unit obtains breast ultrasound image data by using an ultrasound probe and an imaging device, ensuring that the collected image information is accurate and complete and providing original image data for subsequent processing.

[0008] Preferably, the breast ultrasound image preprocessing module performs preprocessing operations such as filtering and noise reduction, contrast enhancement, and image edge preservation, ensuring the improvement of image quality and providing clear and accurate breast ultrasound image data for subsequent three-dimensional reconstruction and lesion detection.

[0009] Preferably, the 3D reconstruction module includes an image registration unit. The image registration unit compares ultrasonic images obtained from different perspectives and time points through precise image registration techniques, adjusts the positions and angles of the ultrasonic images, ensures that the images can be accurately superimposed, prepares for 3D modeling, and guarantees the accuracy and consistency of subsequent 3D models.

[0010] Preferably, the 3D reconstruction module includes a 3D modeling unit. The 3D modeling unit proposes a 3D modeling algorithm for breast ultrasonic images based on tomography. By integrating the registered two-dimensional image data, a 3D model of breast tissue is constructed, ensuring the precise construction and efficient rendering of the 3D model, enabling the system to perform a stereoscopic display and precise positioning of breast lesion areas.

[0011] Specifically, the 3D modeling algorithm for breast ultrasonic images based on tomography is as follows: First, ultrasonic image information of the breast is obtained from multiple angles through tomography technology to reconstruct the internal structure of the object, so as to obtain the pressure field information under a given sound source, frequency, and complex-valued sound speed. The specific formula is expressed as:

[0012]

[0013] where U represents the pressure field, l represents the Laplace operator, C represents the complex-valued sound speed, Δ(Ω) represents the sound source term, Ω represents the angular frequency, and the specific formula is expressed as:

[0014] Ω = 2πf

[0015] where f represents the frequency. By simulating the propagation of ultrasonic waves in three-dimensional space, it lays a foundation for the subsequent reconstruction of the sound speed distribution. Then, the finite difference method is used to divide space and time into small, finite intervals, and then derivatives are solved on these intervals. By discretizing the continuous partial differential equation into a set of algebraic equations that can be solved, the propagation of waves in three-dimensional space is simulated, and the sound source term at the corresponding grid points is obtained by calculating the pressure distribution under a given sound source and medium properties. The specific formula is expressed as:

[0016]

[0017] where Δ(Ω) i,j,k represents the sound source term at the grid point (i, j, k), U i,j,k represents the pressure field value at the grid point (i, j, k), Δx 2 、Δy 2 、Δz 2are respectively represented as the spatial step sizes along the x, y, and z directions. x, y, and z are respectively represented as the breast ultrasound information in the horizontal direction, the breast ultrasound information in the vertical direction, and the breast ultrasound information in the depth direction. i, j, and k are respectively represented as the direction indices along the horizontal direction, the vertical direction, and the depth direction, which are used to access and calculate the values of each point in the three-dimensional space grid. Each index represents a corresponding grid point, jointly defining the position in the three-dimensional space. Secondly, after obtaining the initial pressure field, a cost function E is defined to measure the difference between the simulated wave field and the observed wave field, and the sound speed C is iteratively updated to minimize the cost function. The specific formula of the cost function is as follows:

[0018]

[0019] Among them, E(Ω, C) is represented as the cost function, e is represented as the residual, H is represented as the conjugate operation, and the specific calculation formula of e(Ω, C) is represented as:

[0020] e(Ω, C) = U obs (Ω, C) - D obs (Ω, C true )

[0021] Among them, U obs represents the simulated wave field, D obs represents the observed wave field, C true represents the true sound speed. The iterative update formula of the sound speed is represented as:

[0022]

[0023] Among them, C (p+1) represents the sound speed at the (p + 1)-th iteration, C (p) represents the sound speed at the p-th iteration, and α represents the iterative update step size, represents the gradient of the cost function. By iteratively updating the sound speed, the optimal sound speed distribution can be obtained, thereby realizing the three-dimensional modeling of breast tissue. Secondly, after iteratively updating the sound speed and obtaining the optimal sound speed distribution, three-dimensional reconstruction of breast ultrasound images is performed to obtain the three-dimensional sound speed map of breast tissue. Using the optimized sound speed model for three-dimensional wave field simulation, the propagation of sound waves is evaluated at each grid point, thereby obtaining the detailed three-dimensional structure of breast tissue. The specific formula is represented as:[[ID=]]

[0024]

[0025] Among them, C opt / / There seems to be an incomplete formula here in the original text. Maybe it should be something like "C opt represents the sound speed optimized through the iterative update of the sound speed, G(i, j, k; Ω, C opt " represents the sound speed optimized through the iterative update of the sound speed, G(i, j, k; Ω, C opt) It is expressed as the Green's function in three-dimensional space, representing the wave field at the grid point (i, j, k) due to a point source at the origin. Considering the attenuation and phase change of the wave propagating from the sound source to the receiving point, N is expressed as the number of grid points along each spatial dimension. Finally, after completing the three-dimensional reconstruction, by comparing with the known sound speed distribution, the robustness and accuracy of the algorithm in processing actual breast ultrasound images are ensured. The specific formula is expressed as:

[0026]

[0027] Among them, E final is expressed as the final cost function value, and e final is expressed as the final residual. By using the finite difference method to solve the three-dimensional equation and iteratively optimizing the sound speed model, the accurate three-dimensional reconstruction of breast tissue is achieved. After simulating the propagation of breast ultrasound waves in three-dimensional space, the simulated and actual observed wave fields are compared, and the sound speed distribution is continuously updated until they match, realizing the accurate three-dimensional reconstruction of breast tissue.

[0028] Preferably, the lesion detection and classification module includes a feature extraction unit. The feature extraction unit extracts key features helpful for lesion recognition by analyzing the attributes of shape, texture, and signal distribution in the ultrasound image, ensuring sufficient decision-making information for the subsequent classification and recognition of breast lesions.

[0029] Preferably, the lesion detection and classification module includes a classification and recognition unit. The classification and recognition unit proposes a breast lesion detection algorithm based on a fuzzy neural network. By reducing artifacts in the ultrasound image and extracting rich local detail features and global feature information of the image, the breast image is classified and recognized for lesions, ensuring the accurate recognition and classification of the breast ultrasound image by the system.

[0030] Specifically, the breast lesion detection algorithm based on a fuzzy neural network is as follows: First, the key features of the breast ultrasound image obtained by the feature extraction unit are further extracted for the subsequent input of the fuzzy neural network. The co-occurrence matrix of the key feature region is calculated, and based on this matrix, the entropy information, uniformity information, contrast information, and maximum co-occurrence matrix element of the key features of the breast ultrasound image are extracted. The calculation formula for the co-occurrence matrix element is expressed as:

[0031]

[0032] Among them, P represents the co-occurrence matrix, which represents the matrix of the co-occurrence frequency of gray levels at a specific distance d and angle θ between pixels in the breast ultrasound image. d represents the distance between pixels, which represents the distance between pixel pairs considered when calculating the co-occurrence matrix in the breast ultrasound image. θ represents the angle, which represents the angle between pixel pairs considered when calculating the co-occurrence matrix in the breast ultrasound image. a and b represent the gray values of two pixel points in the breast ultrasound image. f represents the gray function. k represents the row number of the position coordinates in the breast ultrasound image. l represents the column number of the position coordinates in the breast ultrasound image. D represents the set of pixel points when calculating the co-occurrence matrix. δ represents the Kronecker function. cos represents the cosine function. sin represents the sine function. The specific formula for entropy calculation is expressed as:

[0033]

[0034] Among them, H represents the entropy information, which represents the value finally calculated to measure the uncertainty of the regional information in the breast ultrasound image. i, j represent the indices of the rows and columns in the co-occurrence matrix, and p ij represents the probability that gray levels i and j appear simultaneously in the co-occurrence matrix. log represents the logarithmic function. The entropy information can quantify the distribution uniformity of pixel intensity values in the image region. The specific formula for the uniformity information is expressed as:

[0035]

[0036] Among them, U represents the uniformity information of the breast ultrasound image, which represents the value finally calculated to measure the consistency of pixel intensities in the image region. The specific formula for the contrast information of the breast ultrasound image is expressed as:

[0037]

[0038] Regions with large pixel intensity changes are identified by calculating the contrast information of the breast ultrasound image. The specific formula for the maximum co-occurrence matrix element is expressed as:

[0039]

[0040] Among them, MCM represents the maximum co-occurrence matrix element. Then, a fuzzy neural network is constructed using the extracted features of the breast ultrasound image, including entropy information, uniformity information, contrast information, and maximum co-occurrence matrix element information as the input of the fuzzy neural network. The specific formula for the input layer of the fuzzy neural network is expressed as:

[0041] I = [ΔH, ΔU, ΔC, ΔMCM]

[0042] Among them, I represents the input vector, which contains the extracted feature difference values of breast ultrasound images. ΔH represents the entropy difference, ΔU represents the uniformity difference, ΔC represents the contrast difference, and ΔMCM represents the difference of the maximum co-occurrence matrix elements. The specific formula of the fuzzification layer is expressed as:

[0043]

[0044] Among them, O k represents the fuzzification output of the k-th feature, k represents the index of the number of breast ultrasound image features, and I k represents the input value of the k-th feature of the organism. min(I k ) represents the minimum value among all the input values of the k-th feature, and max(I k ) represents the maximum value among all the input values of the k-th feature. The specific formula of the maximum fuzzy neuron layer is expressed as:

[0045]

[0046] Among them, S zk represents the output of the z-th neuron for the k-th input feature. z represents the index of the number of neurons in the fuzzy neural network. μ zk represents the membership function, which is used to measure the membership degree of the input O k in the z-th fuzzy neuron. O k represents the fuzzification output of the k-th feature, c z represents the center value of the z-th neuron, which represents the input value when the neuron is most active. σ z represents the standard deviation of the z-th neuron, which controls the width of the membership function. The specific formula of the fuzzy decision layer is expressed as:

[0047]

[0048] Among them, D v represents the maximum output value of the v-th type of breast lesion type. The breast lesion types include two types: benign and malignant. The specific formula of the final output layer is expressed as:

[0049]

[0050] Among them, Y represents the final classification result, D1 represents the maximum output value of the malignant category, D2 represents the maximum output value of the benign category, and otherwise represents a conditional statement. The constructed fuzzy neural network can use the extracted breast feature differences to detect breast lesions, effectively detect and distinguish malignant masses and benign tissues. The method based on the fuzzy neural network has significant advantages in dealing with uncertainty and nonlinear problems and can effectively and accurately detect breast lesions.

[0051] Preferably, the data storage and management module stores the ultrasonic image data and detection results collected and processed by the system by constructing and maintaining a database, ensuring the integrity, security, and efficient retrieval of the data.

[0052] Preferably, the report generation and display interaction module automatically generates a detection report and provides a user interface, enabling the user to view and analyze the detection results in real time, ensuring the accurate transmission of detection information and the convenience of user operations.

[0053] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0054] The three-dimensional modeling unit proposes a three-dimensional modeling algorithm for breast ultrasound images based on tomography. This algorithm first uses tomography technology to obtain ultrasonic image information of the internal structure of the breast from multiple angles. By simulating the propagation of ultrasonic waves in three-dimensional space, the pressure field distribution under given sound sources, frequencies, and complex sound speeds is calculated. In this process, space and time are discretized by the finite difference method, and partial differential equations are transformed into solvable algebraic equations to simulate the propagation of ultrasonic waves in the three-dimensional space of the breast, thereby accurately estimating the sound source term at each grid point. The gradually generated pressure field distribution provides a solid foundation for the reconstruction of the sound speed distribution. Next, the algorithm defines a cost function to measure the difference between the simulated wave field and the observed wave field, and iteratively updates the sound speed distribution to minimize this cost function, thereby obtaining the optimal sound speed distribution. The optimized sound speed model not only effectively simulates the propagation of breast ultrasonic waves but also provides accurate data support for three-dimensional reconstruction. After obtaining the best sound speed distribution, the acoustic wave propagation is further evaluated at each grid point through the Green's function to achieve a detailed three-dimensional modeling of breast tissue. In addition, by comparing with actual observation data, the robustness and high precision of the three-dimensional modeling are ensured. This process can iteratively optimize the model to reduce the residual and optimize the final cost function value, making the reconstruction result more consistent with the actual structure. In summary, based on this three-dimensional reconstruction model, the system can evaluate the location and nature of breast lesions at higher resolution and accuracy, improving the accuracy and efficiency of breast lesion detection. The high-precision three-dimensional modeling achieved by this algorithm provides strong support for the early detection and treatment plan of breast diseases.

[0055] 1. The classification and recognition unit proposes a breast lesion detection algorithm based on a fuzzy neural network. Through in-depth analysis of the features of breast ultrasound images, effective lesion detection is achieved. By using the key features of breast ultrasound images obtained from the feature extraction unit, the co-occurrence matrix is calculated, and in-depth feature information such as entropy, uniformity, contrast, and the maximum co-occurrence matrix element is extracted from it. The construction of the co-occurrence matrix reflects the spatial relationship between pixels and captures local and global information of image details. Entropy information measures the uncertainty of pixel intensities in an image region, uniformity information provides a measure of pixel consistency, contrast reveals the range of intensity changes in an image region, and the maximum co-occurrence matrix element helps determine the significant feature points of the image. The calculated eigenvalues are normalized through the input layer of the fuzzy neural network, and an input vector is constructed and fuzzified to capture the differences in eigenvalues. The fuzzy neuron layer uses membership functions to fuzzify each input feature to reflect the activity levels of different features to the greatest extent. Through the fuzzy decision layer, the classification results of benign and malignant lesions are finally output. The key of the breast lesion detection algorithm based on the fuzzy neural network lies in the processing ability of the fuzzy neural network for feature uncertainty, which is particularly suitable for fuzzy and non-linear feature data in breast ultrasound images, enabling the system to significantly improve the detection accuracy and classification effect, facilitating effective detection and differentiation between malignant and benign lesions. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The invention will be further described with reference to the accompanying drawings. However, the embodiments in the drawings do not constitute any limitation to the invention. For those of ordinary skill in the art, other drawings can also be obtained based on the following drawings without creative efforts.

[0057] Figure 1 It is a schematic structural diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0058] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0059] Please refer to Figure 1, the present invention provides a three-dimensional reconstruction and lesion detection system for breast ultrasound images, including an ultrasound data acquisition unit, a breast ultrasound image preprocessing module, a three-dimensional reconstruction module, a lesion detection and classification module, a data storage and management module, and a report generation and display interaction module. It is characterized in that: the ultrasound data acquisition unit is used to collect breast ultrasound image data, ensuring the acquisition of high-quality original image information and providing a basis for subsequent processing; the breast ultrasound image preprocessing module is used to perform preprocessing operations such as filtering and noise reduction, contrast enhancement, and image edge preservation on the collected ultrasound images, ensuring image clarity and quality and guaranteeing the accuracy of subsequent lesion detection; the three-dimensional reconstruction module includes an image registration unit and a three-dimensional modeling unit. The image registration unit is used to spatially align ultrasound images obtained from different perspectives and time points, ensuring consistency and comparability between images and providing accurate image data for three-dimensional modeling. The three-dimensional modeling unit proposes a three-dimensional modeling algorithm for breast ultrasound images based on tomography to construct a three-dimensional model according to the registered two-dimensional ultrasound images, ensuring that the three-dimensional structures of breast tissues and lesions can be accurately presented and analyzed; the lesion detection and classification module includes a feature extraction unit and a classification and recognition unit. The feature extraction unit is used to extract key information from breast ultrasound images, including breast morphology, texture, and boundary features, ensuring rich local and global feature data for classification. The classification and recognition unit proposes a breast lesion detection algorithm based on a fuzzy neural network to classify and recognize breast lesions according to the extracted features, ensuring accurate detection and effective differentiation of breast lesions by the system; the data storage and management module is used to store and manage the ultrasound image data and detection reports of patients, ensuring the security, reliability, and efficient access of the data; the report generation and display interaction module is used to generate detection reports and provide a user interaction interface, enabling users to view reports and images in real time, ensuring the accurate transmission of detection results and the convenience of user operations.

[0060] Refer to Figure 1 , further, the ultrasound data acquisition unit obtains breast ultrasound image data by using an ultrasound probe and an imaging device, ensuring that the collected image information is accurate and complete and providing original image data for subsequent processing.

[0061] Refer to Figure 1 , further, the breast ultrasound image preprocessing module improves the image quality by performing preprocessing operations such as filtering and noise reduction, contrast enhancement, and image edge preservation, providing clear and accurate breast ultrasound image data for subsequent three-dimensional reconstruction and lesion detection.

[0062] Refer to Figure 1, Further, the 3D reconstruction module includes an image registration unit. The image registration unit compares ultrasonic images obtained from different perspectives and time points through precise image registration technology, adjusts the positions and angles of the ultrasonic images, ensures that the images can be accurately superimposed, prepares for 3D modeling, and guarantees the accuracy and consistency of the subsequent 3D model.

[0063] Refer to Figure 1 , Further, the 3D reconstruction module includes a 3D modeling unit. The 3D modeling unit proposes a 3D modeling algorithm for breast ultrasonic images based on tomography. By integrating the registered two-dimensional image data, a 3D model of breast tissue is constructed, ensuring the precise construction and efficient rendering of the 3D model, enabling the system to perform a three-dimensional display and precise positioning of breast lesion areas.

[0064] Refer to Figure 1 , Further, the 3D modeling algorithm for breast ultrasonic images based on tomography is as follows: First, ultrasonic image information of the breast is obtained from multiple angles through tomography technology to reconstruct the internal structure of the object, so as to obtain the pressure field information under a given sound source, frequency, and complex-valued sound speed. The specific formula is expressed as:

[0065]

[0066] Among them, U represents the pressure field, l represents the Laplace operator, C represents the complex-valued sound speed, Δ(Ω) represents the sound source term, Ω represents the angular frequency, and the specific formula is expressed as:

[0067] Ω = 2πf

[0068] Among them, f represents the frequency. By simulating the propagation of ultrasonic waves in three-dimensional space, it lays a foundation for the subsequent reconstruction of the sound speed distribution. Then, the finite difference method is used to divide space and time into small, finite intervals, and then derivatives are solved on these intervals. By discretizing the continuous partial differential equation into a set of algebraic equations that can be solved, the propagation of waves in three-dimensional space is simulated, and the sound source term at the corresponding grid points is obtained by calculating the pressure distribution under a given sound source and medium properties. The specific formula is expressed as:

[0069]

[0070] Among them, Δ(Ω) i,j,k represents the sound source term at the grid point (i, j, k), U i,j,k is the pressure field value at the grid point (i, j, k), Δx 2 , Δy 2 , Δz 2are respectively represented as the spatial step sizes along the x, y, and z directions. x, y, and z are respectively represented as the breast ultrasound information in the horizontal direction, the breast ultrasound information in the vertical direction, and the breast ultrasound information in the depth direction. i, j, and k are respectively represented as the direction indices along the horizontal direction, the vertical direction, and the depth direction, which are used to access and calculate the values of each point in the three-dimensional space grid. Each index represents a corresponding grid point, jointly defining the position in the three-dimensional space. Secondly, after obtaining the initial pressure field, a cost function E is defined to measure the difference between the simulated wave field and the observed wave field, and the sound speed C is iteratively updated to minimize the cost function. The specific formula of the cost function is as follows:

[0071]

[0072] Among them, E(Ω, C) is represented as the cost function, e is represented as the residual, H is represented as the conjugate operation, and the specific calculation formula of e(Ω, C) is represented as:

[0073] e(Ω, C) = U obs (Ω, C) - D obs (Ω, C true )

[0074] Among them, U obs is represented as the simulated wave field, D obs is represented as the observed wave field, C true is represented as the true sound speed. The iterative update formula of the sound speed is represented as:

[0075]

[0076] Among them, C (p+1) is represented as the sound speed at the (p + 1)-th iteration, C (p) is represented as the sound speed at the p-th iteration, and α is represented as the iterative update step size. is represented as the gradient of the cost function. By iteratively updating the sound speed, the optimal sound speed distribution can be obtained, thereby realizing the three-dimensional modeling of breast tissue. Secondly, after iteratively updating the sound speed and obtaining the optimal sound speed distribution, three-dimensional reconstruction of breast ultrasound images is performed to obtain the three-dimensional sound speed map of breast tissue. Using the optimized sound speed model for three-dimensional wave field simulation, the propagation of sound waves is evaluated at each grid point, thereby obtaining the detailed three-dimensional structure of breast tissue. The specific formula is represented as:

[0077]

[0078] Among them, C opt is represented as the sound speed optimized by the iterative update of the sound speed, G(i, j, k; Ω, C opt) is represented as the Green's function in three-dimensional space, representing the wave field at the grid point (i, j, k) due to a point source at the origin, considering the attenuation and phase change of the wave propagating from the sound source to the receiving point. N is the number of grid points along each spatial dimension. Finally, after completing the three-dimensional reconstruction, by comparing with the known sound speed distribution, the robustness and accuracy of the algorithm in processing actual breast ultrasound images are ensured. The specific formula is expressed as:

[0079]

[0080] where, E final is represented as the final cost function value, and e final is represented as the final residual. By using the finite difference method to solve the three-dimensional equation and iteratively optimizing the sound speed model, the accurate three-dimensional reconstruction of breast tissue is achieved. After simulating the propagation of breast ultrasound waves in three-dimensional space, the simulated and actual observed wave fields are compared, and the sound speed distribution is continuously updated until they match, realizing the accurate three-dimensional reconstruction of breast tissue.

[0081] Refer to Figure 1 , furthermore, the lesion detection and classification module includes a feature extraction unit. The feature extraction unit extracts key features that contribute to lesion recognition by analyzing the attributes of shape, texture, and signal distribution in the ultrasound image, ensuring sufficient decision-making information for the subsequent classification and recognition of breast lesions.

[0082] Refer to Figure 1 , furthermore, the lesion detection and classification module includes a classification and recognition unit. The classification and recognition unit proposes a breast lesion detection algorithm based on a fuzzy neural network, classifies and recognizes breast lesions in breast images by reducing artifacts in the ultrasound image and extracting rich local detail features and global feature information of the image, ensuring the accurate recognition and classification of the system for breast ultrasound images.

[0083] Refer to Figure 1 , furthermore, the breast lesion detection algorithm based on the fuzzy neural network is specifically as follows: First, the key features of the breast ultrasound image obtained by the feature extraction unit are further extracted for the subsequent input of the fuzzy neural network. Calculate the co-occurrence matrix of the key feature region, and based on this matrix, extract the entropy information, uniformity information, contrast information, and the maximum co-occurrence matrix element of the key features of the breast ultrasound image. The calculation formula for the co-occurrence matrix element is expressed as:

[0084]

[0085] Among them, P represents the co-occurrence matrix, which represents the matrix of the co-occurrence frequency of gray levels at a specific distance d and angle θ between pixels in the breast ultrasound image. d represents the distance between pixels, which represents the distance between pixel pairs considered when calculating the co-occurrence matrix in the breast ultrasound image. θ represents the angle, which represents the angle between pixel pairs considered when calculating the co-occurrence matrix in the breast ultrasound image. a and b represent the gray values of two pixel points in the breast ultrasound image. f represents the gray function. k represents the row number of the position coordinates in the breast ultrasound image. l represents the column number of the position coordinates in the breast ultrasound image. D represents the set of pixel points when calculating the co-occurrence matrix. δ represents the Kronecker function. cos represents the cosine function. sin represents the sine function. The specific formula for entropy calculation is expressed as:

[0086]

[0087] Among them, H represents the entropy information, which represents the value finally calculated to measure the uncertainty of the regional information of the breast ultrasound image. i, j represent the indexes of the rows and columns in the co-occurrence matrix, and p ij represents the probability that gray levels i and j appear simultaneously in the co-occurrence matrix. log represents the logarithmic function. The entropy information can quantify the distribution uniformity of pixel intensity values in the image region. The specific formula for the uniformity information is expressed as:

[0088]

[0089] Among them, U represents the uniformity information of the breast ultrasound image, which represents the value finally calculated to measure the consistency of pixel intensities in the image region. The specific formula for the contrast information of the breast ultrasound image is expressed as:

[0090]

[0091] By calculating the contrast information of the breast ultrasound image, the regions with large pixel intensity changes are identified. The specific formula for the maximum co-occurrence matrix element is expressed as:

[0092]

[0093] Among them, MCM represents the maximum co-occurrence matrix element. Then, a fuzzy neural network is constructed using the extracted features of the breast ultrasound image, including entropy information, uniformity information, contrast information, and maximum co-occurrence matrix element information as the input of the fuzzy neural network. The specific formula for the input layer of the fuzzy neural network is expressed as:

[0094] I = [ΔH, ΔU, ΔC, ΔMCM]

[0095] Among them, I represents the input vector, which contains the extracted differential values of breast ultrasound image features. ΔH represents the entropy difference, ΔU represents the uniformity difference, ΔC represents the contrast difference, and ΔMCM represents the difference of the maximum co-occurrence matrix elements. The specific formula of the fuzzification layer is expressed as:

[0096]

[0097] Among them, O k represents the fuzzification output of the k-th feature, k represents the index of the number of breast ultrasound image features, and I k represents the input value of the k-th feature of the organism. min(I k ) represents the minimum value among all input values of the k-th feature, and max(I k ) represents the maximum value among all input values of the k-th feature. The specific formula of the maximum fuzzy neuron layer is expressed as:

[0098]

[0099] Among them, S zk represents the output of the z-th neuron for the k-th input feature. z represents the index of the number of neurons in the fuzzy neural network. μ zk represents the membership function, which is used to measure the membership degree of the input O k in the z-th fuzzy neuron. O k represents the fuzzification output of the k-th feature, and σ z represents the center value of the z-th neuron, which represents the input value when the neuron is most active. σ z represents the standard deviation of the z-th neuron, which controls the width of the membership function. The specific formula of the fuzzy decision layer is expressed as:

[0100]

[0101] Among them, D v represents the maximum output value of the v-th type of breast lesion. The types of breast lesions include two types: benign and malignant. The specific formula of the final output layer is expressed as:

[0102]

[0103] Among them, Y represents the final classification result. D1 represents the maximum output value of the malignant category, D2 represents the maximum output value of the benign category, and otherwise represents a conditional statement. The constructed fuzzy neural network can use the extracted breast feature differences to detect breast lesions, effectively detect and distinguish malignant masses and benign tissues. The method based on the fuzzy neural network has significant advantages in dealing with uncertainty and nonlinear problems and can effectively and accurately detect breast lesions.

[0104] Refer to Figure 1 Furthermore, the data storage and management module stores the ultrasonic image data and detection results collected and processed by the system by constructing and maintaining a database, ensuring the integrity, security, and efficient retrieval of the data.

[0105] Refer to Figure 1 Furthermore, the report generation and display interaction module automatically generates a detection report and provides a user interface, enabling the user to view and analyze the detection results in real time, ensuring the accurate conveyance of detection information and the convenience of user operations.

[0106] In specific use, first, the ultrasonic data acquisition unit is used to collect breast ultrasonic image data, ensuring the acquisition of high-quality original image information as the basis for subsequent processing. Second, the breast ultrasonic image preprocessing module performs preprocessing operations such as filtering and noise reduction, contrast enhancement, and image edge preservation on the collected ultrasonic images, ensuring image clarity and quality and guaranteeing the accuracy of subsequent lesion detection. Then, the 3D reconstruction module includes an image registration unit and a 3D modeling unit. The image registration unit is used to spatially align ultrasonic images obtained from different perspectives and time points, ensuring consistency and comparability between images and providing accurate image data for 3D modeling. The 3D modeling unit proposes a 3D modeling algorithm for breast ultrasonic images based on tomography to construct a 3D model according to the registered 2D ultrasonic images, ensuring that the three-dimensional structure of breast tissue and lesions can be accurately presented and analyzed. Second, the lesion detection and classification module includes a feature extraction unit and a classification and recognition unit. The feature extraction unit is used to extract key information from breast ultrasonic images, including breast morphology, texture, and boundary features, ensuring rich local and global feature data for classification. The classification and recognition unit proposes a breast lesion detection algorithm based on a fuzzy neural network to classify and recognize breast lesions according to the extracted features, ensuring the accurate detection and effective discrimination of breast lesions by the system. Finally, the data storage and management module is used to store and manage the ultrasonic image data and detection reports of patients, ensuring the security, reliability, and efficient access of the data. The report generation and display interaction module is used to generate a detection report and provide a user interaction interface, enabling the user to view the report and images in real time, ensuring the accurate conveyance of detection results and the convenience of user operations.

[0107] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A system for three-dimensional reconstruction and lesion detection of breast ultrasound images, characterized in that: It includes an ultrasound data acquisition unit, a breast ultrasound image preprocessing module, a three-dimensional reconstruction module, a lesion detection and classification module, a data storage and management module, and a report generation and display interaction module; wherein the ultrasound data acquisition unit is used to collect breast ultrasound image data, obtain high-quality original image information, and provide a basis for subsequent processing; the breast ultrasound image preprocessing module is used to perform preprocessing operations such as filtering and noise reduction, contrast enhancement, and image edge preservation on the collected ultrasound images; the three-dimensional reconstruction module includes an image registration unit and a three-dimensional modeling unit; the image registration unit is used to spatially align ultrasound images acquired from different perspectives and time points; the three-dimensional modeling unit is used to construct a three-dimensional model from the registered two-dimensional ultrasound image according to the three-dimensional modeling algorithm of breast ultrasound images based on tomography, so that the three-dimensional structure of breast tissue and lesions can be accurately presented and analyzed; the lesion detection and classification module includes a feature extraction unit and a classification recognition unit; the feature extraction unit is used to extract key information from the breast ultrasound image, including breast morphology, texture and boundary features, to provide rich local and global feature data for classification; the classification recognition The identification unit is based on a fuzzy neural network breast lesion detection algorithm, which is used to classify and identify breast lesions according to the extracted features; the data storage and management module is used to store and manage the patient's ultrasound image data and test reports; the report generation and display interaction module is used to generate a test report and provide a user interaction interface; the three-dimensional reconstruction module includes an image registration unit and a three-dimensional modeling unit. The image registration unit uses precise image registration technology to compare ultrasound images acquired from different perspectives and time points, adjust the position and angle of the ultrasound image, and accurately superimpose the images to prepare for three-dimensional modeling and ensure the accuracy and consistency of subsequent three-dimensional models; the three-dimensional modeling unit proposes a three-dimensional modeling algorithm for breast ultrasound images based on tomography. By integrating the registered two-dimensional image data, a three-dimensional model of breast tissue is constructed. The precise construction and efficient rendering of the three-dimensional model enable the system to stereoscopically display and accurately locate the breast lesion area. First, breast ultrasound image information is acquired from multiple angles through tomography technology, and the internal structure of the object is reconstructed to obtain pressure field information under a given sound source, frequency and complex-valued sound velocity. The specific formula is expressed as follows: Among them, U represents the pressure field, It is represented by the Laplace operator, C represents the complex-valued sound speed, Δ(Ω) represents the sound source term, and Ω represents the angular frequency. The specific formula is expressed as follows: Ω=2πf Among them, f represents the frequency. By simulating the propagation of ultrasonic waves in three-dimensional space, the foundation is laid for the subsequent reconstruction of the sound velocity distribution. Then, the finite difference method is used to divide space and time into small, finite intervals. The derivatives are then solved on these intervals. The continuous partial differential equations are discretized into a set of solvable algebraic equations to simulate the propagation of waves in three-dimensional space. The sound source term at the corresponding grid point is obtained by calculating the pressure distribution under given sound source and medium properties. The specific formula is expressed as follows: Where, Δ(Ω) i,j,k Expressed as the sound source term at the grid point (i, j, k), U i,j,k The pressure field value at the grid point (i, j, k), Δx 2 , Δy 2 , Δz 2 They are represented as the spatial step lengths along the x, y, and z directions, respectively. x, y, and z represent the horizontal breast ultrasound information, vertical breast ultrasound information, and depth breast ultrasound information, respectively. i, j, and k represent the direction indexes along the horizontal, vertical, and depth directions, respectively. They are used to access and calculate the value of each point in the three-dimensional space grid. Each index represents a corresponding grid point, which together define the position in the three-dimensional space. Secondly, after obtaining the initial pressure field, a cost function E is defined to measure the difference between the simulated wave field and the observed wave field, and the sound speed C is iteratively updated to minimize the cost function. The specific formula of the cost function is: Among them, E(Ω, C) represents the cost function, e represents the residual, H represents the conjugate operation, and the specific calculation formula of e(Ω, C) is expressed as: e(Ω,C)=U obs (Ω,C)-D obs (Ω,C true ) Among them, U obs Expressed as simulated wave field, D obs Expressed as the observed wave field, C true Expressed as the true speed of sound, the iterative update formula of the speed of sound is expressed as: Among them, C (p+1) Expressed as the speed of sound at the p+1th iteration, C (p) It represents the speed of sound at the pth iteration, α represents the iterative update step size, It is expressed as the gradient of the cost function. By iteratively updating the sound velocity, the optimal sound velocity distribution can be obtained, thereby realizing the three-dimensional modeling of breast tissue. Secondly, after iteratively updating the sound velocity and obtaining the optimal sound velocity distribution, three-dimensional reconstruction of the breast ultrasound image is performed to obtain a three-dimensional sound velocity map of the breast tissue. The optimized sound velocity model is used to perform three-dimensional wave field simulation, and the propagation of the sound wave is evaluated at each grid point to obtain the detailed three-dimensional structure of the breast tissue. The specific formula is expressed as follows: Among them, C opt It is expressed as the sound speed after iterative update optimization, G(i,j,k;Ω,C opt ) is expressed as a Green's function in three-dimensional space, representing the wave field generated by the point source at the origin at the grid point (i, j, k). Considering the attenuation and phase change of the wave propagating from the sound source to the receiving point, N is expressed as the number of grid points along each spatial dimension. The relationship between the final residual and the final cost function value is constructed as follows: Among them, E final Expressed as the final cost function value, e final Expressed as the final residual, the finite difference method is used to solve the three-dimensional equations and the sound velocity model is iteratively optimized to achieve accurate three-dimensional reconstruction of breast tissue. After simulating the propagation of breast ultrasound in three-dimensional space, the simulated and actual observed wave fields are compared, and the sound velocity distribution is continuously updated until a match is achieved to achieve accurate three-dimensional reconstruction of breast tissue.

2. The system for 3D reconstruction and lesion detection of breast ultrasound images according to claim 1, characterized in that: The lesion detection and classification module includes a feature extraction unit and a classification and recognition unit; the feature extraction unit extracts key features that are helpful for lesion identification by analyzing the properties of shape, texture and signal distribution in the ultrasound image, providing sufficient decision-making information for the subsequent classification and recognition of breast lesions; the classification and recognition unit is based on a breast lesion detection algorithm of a fuzzy neural network, which classifies and recognizes lesions in the breast image by reducing artifacts in the ultrasound image and extracting rich local detail features and global feature information of the image; first, the key features of the breast ultrasound image obtained by the feature extraction unit are further extracted for subsequent fuzzy neural network input, the co-occurrence matrix of the key feature area is calculated, and based on this matrix, the entropy information, uniformity information, contrast information and maximum co-occurrence matrix elements of the key features of the breast ultrasound image are extracted. The calculation formula of the co-occurrence matrix elements is expressed as: Where P is the co-occurrence matrix, which represents the matrix of grayscale co-occurrence frequencies at a specific distance d and angle θ between pixels in the breast ultrasound image. d is the distance between pixels, which represents the distance between pixel pairs considered when calculating the co-occurrence matrix in the breast ultrasound image. θ is the angle, which represents the angle between pixel pairs considered when calculating the co-occurrence matrix in the breast ultrasound image. a and b represent the grayscale values of two pixels in the breast ultrasound image. f represents the grayscale function. It is represented as the row number of the position coordinate in the breast ultrasound image, It represents the column number of the position coordinate in the breast ultrasound image, D represents the set of pixel points when calculating the co-occurrence matrix, δ represents the Kronecker function, cos represents the cosine function, and sin represents the sine function. The specific formula for entropy calculation is expressed as follows: Among them, H represents entropy information, which represents the final calculated value to measure the uncertainty of regional information in breast ultrasound images. Represented as gray levels in the co-occurrence matrix and grayscale The probability of simultaneous occurrence, log, is expressed as a logarithmic function. The specific calculation formula of uniformity information is expressed as: in, It is expressed as the uniformity information of the breast ultrasound image, which represents the final calculated value that measures the consistency of pixel intensity in the image area. The specific formula of the contrast information of the breast ultrasound image is expressed as: By calculating the contrast information of breast ultrasound images, the areas with large pixel intensity changes are identified. The specific formula of the maximum co-occurrence matrix element is expressed as: Among them, MCM is expressed as the maximum co-occurrence matrix element. Then, the fuzzy neural network is constructed using the extracted breast ultrasound image features, including entropy information, uniformity information, contrast information and maximum co-occurrence matrix element information as the input of the fuzzy neural network. The specific formula of the fuzzy neural network input layer is expressed as: I=[ΔH, ΔU, ΔC, ΔMCM] Where I represents the input vector, which contains the difference value of the extracted breast ultrasound image features, ΔH represents the entropy difference, ΔU represents the uniformity difference, ΔC represents the contrast difference, and ΔMCM represents the maximum co-occurrence matrix element difference. The specific formula of the fuzzy layer is expressed as follows: in, Expressed as The fuzzy output of the features, Represented as the index of the number of breast ultrasound image features, Represented as biological The input value of the feature, Expressed as The minimum value among all input values of the feature, Expressed as The maximum value of all input values of the feature, the specific formula of the maximum fuzzy neuron layer is expressed as: in, It is represented by the zth neuron to the The output of the input features, z represents the index of the number of neurons in the fuzzy neural network, Expressed as membership function, c z It is expressed as the center value of the z-th neuron, representing the input value when the neuron is most active, σ z Expressed as the standard deviation of the z-th neuron, it controls the width of the membership function. The specific formula of the fuzzy decision layer is expressed as: Among them, D v It is expressed as the maximum output value of the vth type of breast lesion. Breast lesions include benign and malignant types. The specific formula of the final output layer is expressed as: Among them, Y represents the final classification result, D1 represents the maximum output value of the malignant category, D2 represents the maximum output value of the benign category, and otherwise represents the conditional statement.

3. The system for 3D reconstruction and lesion detection of breast ultrasound images according to claim 1, characterized in that: The ultrasound data acquisition unit acquires ultrasound image data of a breast region by using an ultrasound probe and an imaging device.

4. The system for 3D reconstruction and lesion detection of breast ultrasound images according to claim 1, characterized in that: The breast ultrasound image preprocessing module improves image quality by performing preprocessing operations such as filtering and noise reduction, contrast enhancement, and image edge preservation, thereby providing clear and accurate breast ultrasound image data for subsequent three-dimensional reconstruction and lesion detection.

5. The system for 3D reconstruction and lesion detection of breast ultrasound images according to claim 1, characterized in that: The data storage and management module stores the ultrasound image data and test results collected and processed by the system by building and maintaining a database, ensuring the integrity, security and efficient retrieval of the data.

6. The system for 3D reconstruction and lesion detection of breast ultrasound images according to claim 1, characterized in that: The report generation and display interaction module automatically generates test reports and provides a user interface, allowing users to view and analyze test results in real time, accurately conveying test information and facilitating user operations.

Images